Datasets:

ccdv
/

arxiv-summarization

Dataset card Data Studio Files Files and versions Community

Subset (2)

section · 216k rows

Split (3)

train · 203k rows

article stringlengths 0 745k	abstract stringlengths 0 92.7k
the most massive galaxies in the local universe are thought to have evolved to their current state via a series of dry mergers of relatively gas - poor galaxies over the last 10gyr ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? their ancestors the so - called compact quiescent galaxies ( cqgs ) are found at @xmath10 in sensitive , near - infrared ( nir ) imaging surveys ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? these cqgs have @xmath18@xmath19 smaller effective radii ( @xmath11kpc ) and are @xmath20 denser than their local descendants ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and the process by which they form remains a mystery . recent attempts to probe their star - forming phase using conventional nir observations resulted in the discovery of a relatively unobscured starburst , seen around @xmath213 @xcite . however , detailed simulations and population - synthesis modeling suggest that major mergers at @xmath226 likely play a major role in the formation of the compact stellar component , via dust - obscured compact starbursts ( e.g. * ? ? ? * ; * ? ? ? we must thus penetrate deep within these dusty environments to reveal this vigorous starburst phase . submillimeter galaxies ( smgs ) ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) have long been thought to be plausible progenitors of massive passive galaxies around @xmath232 based on their volume densities @xcite . early source size measurements for @xmath243 smgs using radio continuum and co emission - line data reported a median source size of @xmath25.5 ( full width at half maximum ; fwhm ) corresponding to a radius of @xmath263kpc ( e.g. * ? ? ? * ; * ? ? ? * ) . these early studies resulted in the common notion that high - redshift smgs have larger star - forming regions than luminous , dusty galaxies in the local universe , indicating that the size of their star - forming region is inconsistent with the compact structure of cqgs ( although a few smgs at @xmath16 with compact cores were reported * ? ? ? * ) . the size of starburst regions in smgs at @xmath27 has remained largely unexplored , partly because it is difficult to identify smgs at @xmath27 , partly because the cosmological dimming then makes it difficult to measure their source sizes in the radio regime . there have been a few source size measurements via ( sub)millimeter continuum imaging for @xmath27 smgs ; sma observations of aztec1 at @xmath28 and pdbi observations of hfls3 at @xmath29 revealed radii of @xmath30kpc @xcite . however these two are among the most brightest submillimeter galaxies known ( @xmath31l@xmath3 ) and we need to image more typical smgs with @xmath32l@xmath3 . sensitivity limitations of existing arrays meant we needed to wait for alma in order to measure the far - infrared ( fir ) sizes of @xmath14 smgs for a significant sample of targets . here , we exploit high - resolution continuum imaging with alma to peer within a carefully selected sample of the most distant smgs the most vigorous , dust - obscured , starburst galaxies in the early universe . we demonstrate that they have the compact starburst nuclei necessary to produce the small structures that typify cqgs . we adopt throughout a cosmology with @xmath33kms@xmath5mpc@xmath5 , @xmath34 and @xmath35 . our alma program ( 2012.1.00326.s , p.i.ikarashi ; ikarashi et al.in prep ) was designed to study the most distant dusty starbursts , for which redshift estimates were obtained based on ( sub)millimeter / radio @xcite and red ( sub)millimeter ( e.g. * ? ? ? * ; * ? ? ? * ) colors . in this alma program , we observed 30 aztec sources in the subaru / xmm - newton deep field ( sxdf ) , which includes the ukidss uds field @xcite . the aztec 1100-@xmath36 m map contains a total of 221 millimeter sources over a contiguous area of 950 arcmin@xmath37 . we selected our alma targets based on their faintness in the @xmath38 images ( @xmath39 ; 3@xmath40 ; * ? ? ? * ) and vla 1.4-ghz map ( @xmath4135 @xmath36jy ; 5@xmath40 ) ( arumugam et al . , in preparation ) . with the alma observations we detected 35 significant ( @xmath42 5@xmath40 ) smgs ( hereafter asxdf sources ) associated with 30 aztec sources . given the strong negative k - correction at @xmath438001300@xmath36 m , the faintness in the _ herschel _ and radio bands indicate that these 1100-@xmath36m - selected galaxies are expected to be at high redshifts , i.e. , @xmath14 . note that these galaxies constitute a complementary population to that studied in early sub - millimetre galaxy studies , which were biased towards radio - bright sources , and found to lie at lower redshifts @xmath443 ( e.g. * ? ? ? * ) . among the 35 alma sources , a total of 17 have detections with @xmath45 in the alma 1100-@xmath36 m continuum map . such a high @xmath46 threshold ensures that we can study their sizes with good accuracy with the alma continuum data ( see details in section3.1 ) . as the focus of this paper is on @xmath14 smgs , we analyse here only the 13 ( out of 17 ) sources that have photometric redshifts @xmath47 , or are faint in irac ( @xmath48 22.75 m@xmath49 ) and detected in at most four optical / near - mid - ir broad bands , indicating a likely high redshift , as we explain below . at these shorter wavelengths ( optical through mid - ir ) , we have performed the spectral energy distribution ( sed ) analysis of our sources based on 12 bands , namely @xmath50 , @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 , @xmath57-bands and irac 3.6 , 4.5 , 5.8 and 8.0@xmath36 m ( ikarashi et al.in prep ) , using the same method described in @xcite . for three out of our 13 sources , we obtained redshift estimates , @xmath58 , and derived parameters ( table [ tbl-1 ] ) . the remaining ten sources are only detected in four or less broad bands , so no robust @xmath58 can be obtained from the sed fitting . figure [ fig : z45mic ] shows a 4.5-@xmath36m redshift plot for alma sources and those in the alma - identified smg sample ( aless ) reported by @xcite . the dashed line in this plot indicates the median 4.5-@xmath36m redshift relation , and the solid line corresponds to this same relation minus the @xmath59 scatter , which we have derived using the average sed of the aless sources ( see figure 8 in * ? ? ? we expect more than 85% of the asxdf sources with @xmath60 m@xmath49 to be located at @xmath14 . about 15% of smgs are expected to have @xmath61 fainter than the solid black curve in figure [ fig : z45mic ] at each redshift , and 15% of smgs at @xmath62 are expected to have @xmath60 m@xmath49 . so , by selecting only those galaxies with no redshift estimate in our sample , with @xmath60 , we obtain a conservative list of sources likely to lie at @xmath27 . in table [ tbl-1 ] , we list the expected minimum redshifts for our sources based on the solid line in figure [ fig : z45mic ] . the stacked submillimeter radio sed of these optical / nir dropout smgs also help us to understand , in an independent manner , whether our galaxies are actually located at @xmath14 . figure [ fig : sed ] shows the stacked fluxes at 100 , 160 , 250 , 350 , 500@xmath36 m ( pacs and spire ) , 1100@xmath36 m ( alma ) and 21 cm ( vla ) with the sed of the averaged smgs ( @xmath63k ) at @xmath643 , 4 and 5 . all of the stacked fluxes and errors are based on bootstrapping analysis . we see that the stacked submillimeter / radio fluxes are best fitted at @xmath65 . given a stacked vla 1.4-ghz flux density of @xmath66@xmath36jy and the observed alma 1100-@xmath36 m flux density , we expect a photometric redshift , @xmath67 , for our asxdf sources , based on their radio/(sub)millimeter color ( e.g. * ? ? ? note that , for this exercise , we have considered the radio - fir sed template of the averaged smgs @xcite derived from the alma - identified smgs at @xmath16 @xcite . here we are assuming that this template is also valid at @xmath144 . if @xmath68 follows the trend that smgs at higher redshifts have higher @xmath68 , then the redshifts should indeed be @xmath69 . at the moment , it is difficult to be confident of such a trend as the samples of known @xmath144 smgs are small . nevertheless , from all our arguments based on the multi - wavelength sed study of our galaxies from the optical through the radio , we can conclude that our 13 asxdf / alma sources are safe candidates for @xmath14 smgs . our derived flux densities at 1100-@xmath36 m range from 1.5 to 3.4mjy , corresponding to star formation rates ( sfrs ) of @xmath70600m@xmath3yr@xmath5 and @xmath71@xmath2l@xmath3 . the median sfr and @xmath72 are 340@xmath73m@xmath3yr@xmath5 and 3.4@xmath7410@xmath75l@xmath3 , respectively . the properties of our sample are summarized in table [ tbl-1 ] . the sfrs and @xmath72 are estimated from the average smg seds . we considered uniform redshift probability density at @xmath76 for sources without a @xmath58 determination , and a 1@xmath40 error for sources which do have a @xmath58 . m flux of submillimeter galaxies as a function of redshift . blue dots mark alma - identified laboca ( aless ) sources @xcite . red points mark asxdf sources with photometric redshifts in our sample for source size measurements . black curves show the redshift-4.5-@xmath36 m relation expected from the absolute @xmath56-band flux distribution of aless sources and the optical / nir sed of average aless sources @xcite ; dashed line is for smgs with the median absolute @xmath56-band flux ; solid shows the absolute @xmath56-band flux distribution minus @xmath59 . light red bars mark the 4.5-@xmath36 m flux of the asxdf sources without photometric redshifts . solid green horizontal bar marks @xmath77 at 22.75 m@xmath49 , which is the threshold for selection of @xmath14 smgs in this paper . in this paper , we adopt the cross points between the solid black curve and the horizontal light red lines as the expected 1@xmath40 lower limits of redshift for each source ( these values are listed in table [ tbl-1 ] ) . [ fig : z45mic ] ] . error are estimated by bootstrapping analysis . colored sed is that of average smgs @xcite for @xmath643 , 4 and 5 , as best fit to the alma flux . for pacs 100 and 160 @xmath36 m and spire 250 @xmath36 m data , we plot 3@xmath40 upper limits . [ fig : sed ] ] we measured the source sizes of our asxdf sources using alma continuum data centered at 265ghz . our alma observations were obtained in three blocks , with only small differences in antenna configurations between blocks . seven of the 13 asxdf sources were observed with 25 working 12-m antennas , mainly covering baselines up to 400k@xmath78 , corresponding to physical baseline lengths of 440 m . the remaining six sources were observed with three more 12-m antennas , deployed for tests on longer baselines , covering up to 1200k@xmath78 or 1320 m . the extended - baseline data from 400 to 1200k@xmath78 are used here only as supplementary data because of their limited @xmath79 coverage ( figure [ fig : uvcov ] ) . on - source observation times were 3.64.5 minutes , sufficient to achieve r.m.s . noise levels of 7088@xmath36jybeam@xmath5 . the synthesized beam size in our alma continuum images , using baselines up to 400k@xmath78 , is @xmath80 ( fwhm ) too coarse to allow us to resolve any compact starburst nuclei in high - redshift smgs . all of our sample shows the millimeter sizes of @xmath81 times smaller than the beam size by a casa task , imfit ; about the 5 of the 13 are unresolved or point - like . in this paper we have measured source sizes using the visibility data directly on @xmath79amplitude plots ( hereafter @xmath79-amp plots ) assuming a symmetrical gaussian ) is empirically @xmath82 larger than the size measured ( @xmath83 ) for measurements at @xmath84k@xmath78 . ] as was done in previous studies , to exploit the long - baseline ( @xmath85400k@xmath78 ) data for source size measurements ( figure [ fig : uvamp ] ) . source - size measurements using @xmath79-amp plots have often been made in previous studies using , e.g. , sma and carma , in order to better constrain the size of largely unresolved sources in an image ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? this is equivalent to measuring the circularized effective radius , @xmath8 . in this paper , we have been able to polish this method , owing to the high data quality from alma . we have evaluated the accuracy of our source - size measurements using a monte - carlo simulation , for the purpose of correcting for any systematics and obtaining more reliable source sizes . we generated 82000 mock sources with a symmetric gaussian profile in noisy visibility data , for a range of source sizes and flux densities that cover the putative parameter range of our asxdf sources . we measured source sizes and created cleaned continuum images in the same manner as we had done for our real targets , in order to derive a relation between the input source size , the measured source size and the signal - to - noise ratio in a continuum image . figure [ fig : simu ] shows the derived relation between measured source size found by fitting in @xmath79-amp plots and the ` actual ' size input for the simulation , each versus source size for continuum detections of 10 and 15@xmath40 . this plot demonstrates that our source size measurement is accurate to within 1@xmath40 and that actual source sizes are systematically a little bit larger than the measured source sizes . we therefore adopt source sizes after making a correction based on this relation between measured and actual source sizes ; the correction is done using the probability distribution of actual source size for the appropriate signal - to - noise ratio and measured size of each source . in this paper , we measure millimeter size of asxdf sources with @xmath86 continuum detections . this is because size measurement in the visibility data for 10@xmath40 sources gets less sensitive at fwhm@[email protected] , i.e. , losing inearity ( figure [ fig : simu ] ) . measurement of our sample is safe from this issue ; there is just one asxdf source with s / n@xmath8810 , but its measured size is @[email protected] , and we checked that our measurement for the second lowest s / n of 11.3 , is sensitive down to [email protected] . the measured source sizes are listed in table [ tbl-1 ] . the source sizes of our sample range from [email protected] to [email protected] with a median of [email protected]@xmath7 ( figure [ fig : size ] ) . we also check the dependency of the circularized size measured in @xmath79-amp plots on ellipticity via simulations , for minor / major axis ratio of 0.5 , 0.6 , 0.7 , 0.8 , 0.9 and 1.0 ( figure [ fig : simu ] ) . these simulation suggest that the measured circularized size does not depend strongly on the ellipticity ; however , a weak dependency exists : at an axis ratio of 0.5 , the circularized size can be over - estimated about 10 percent . coverage for asxdf sources with long - baseline antennas for schedule block 1 in table . [ tbl-1 ] . we use visibilities at @xmath79 distances of @xmath84k@xmath78 for source - size measurements where @xmath90@xmath91 coverage is well sampled . we use visibilities at 4001200k@xmath78 only to check the consistency between the expected long - baseline visibilities and the measured size . [ fig : uvcov ] ] as we noted above , six of 13 asxdf sources in our sample were observed with additional three long - baseline antennas covering up to 1200k@xmath78 ( table [ tbl-1 ] ) which enable us to make millimeter images of the sources with an angular resolution of [email protected] ( fwhm ) . using the long - baseline data , we check some concerns in our source size measurements assumptions made about source multiplicity ( i.e. that there is none ) , the possibility of source ellipticity , and the possibility of faint , extended emission . in figure [ fig : uvamp ] , we present high - resolution alma continuum images and @xmath79-amp plots of asxdf1100.013.1 , 27.1 , 45.1 , 45.2 , 49.1 and 53.1 , which are covered by long - baseline data individually . moreover in order to check the properties of fainter sources with better signal - to - noise asxdf1100.027.1 , 045.1 , 045.2 and 049.1 we stacked the visibility data of these four sources using the casa code , stacker @xcite . we also stacked the six sources with long - baseline data to check their average properties . hereafter we refer to the former and latter stacked data as _ stacked faint _ and _ stacked all_. first , @xmath79-amp plots of asxdf1100.013.1 , 27.1 , 45.1 , 45.2 , 49.1 and 53.1 and _ stacked faint _ and _ stacked all _ demonstrate that estimating their size by @xmath79-amp analysis using up to 400k@xmath78 yields results consistent with their long - baseline visibility data up to 1200k@xmath78 ( figure [ fig : uvamp ] ) . second , we created high - angular - resolution alma millimeter continuum images ( hereafter high - res images ) using 2001200k@xmath78 baselines ( figure [ fig : uvamp ] ) . the resultant synthesized beam is [email protected] @xmath92 [email protected] ( pa @xmath88 21@xmath93 ) . the r.m.s . of the images of asxdf1100.053.1 , 13.1 , 27.1 , 45.1 , 45.2 , 49.1 , _ stacked faint _ and _ stacked all _ are 126 , 124 , 124 , 122 , 122 , 126 , 88 and 67 @xmath36jy beam@xmath5 , respectively . the sources are detected with @xmath9411 , 9 , 7 , 7 , 6 , 6 , 10 and 15@xmath40 , respectively . we measured their millimeter sizes and fluxes in the image using a casa task , imfit : [email protected] - [email protected] ( major axis ) . as suggested by our source size measurements via @xmath79-amp fitting , each of our @xmath14 smgs has a compact star - forming region . next , we check whether or not the compact star - forming region dominates the huge sfrs of these smgs . usually we would simply compare the flux of a compact component on a high - res image with the total flux in order to estimate the flux fraction emitted by the compact component . however , we remove the alma data at @xmath79-distances @xmath95k@xmath78 in order to sharpen the synthesized beam , so we need to take into account any missing flux . in this paper , then , we compare fluxes measured using imfit on the high - res alma images with fluxes expected at a @xmath79-distance of @xmath96k@xmath78 . we adopt the fluxes measured at a @xmath79-distance of @xmath97k@xmath78 ( figure [ fig : uvamp ] ) as the flux expected for the measured size . figure [ fig : comp ] shows the comparison between the flux measured at a @xmath79-distance of @xmath96k@xmath78 and flux measured via imfit . the comparison shows that the fluxes measured by imfit are almost the same as the fluxes measured using the visibilities . the relation between fluxes from the image and from the visibility data can be fit by @xmath98 where @xmath99 is the flux measured by imfit and @xmath100 is that measured from the visibilities . this indicates that @xmath89100 percent of the rest - frame fir emission in @xmath14 asxdf sources comes from the compact component shown in the high - res images . last , we check the ellipticity of sources . sizes measured by imfit are shown on the high - res alma images in figure [ fig : uvamp ] , and tend to show ellipticity , i.e. , minor / major axis ratio of a fitted asymmetric gaussian @xmath87 1 ; the minor / major axis ratio of asxdf1100.053.1 , 013.1 , 027.1 , 045.1 and 045.2 are [email protected] , [email protected] , [email protected] , [email protected] , and [email protected] , respectively . here we check the ellipticity shown in the individual sources with @xmath10210 ( figure [ fig : uvamp ] ) . in order to test whether the elliptical feature is real , we checked the empirical accuracy of imfit using monte carlo simulations for major / minor axis ratios of 0.5 , 0.6 , 0.7 , 0.8 , 0.9 and 1.0 and circularized sizes of [email protected] , [email protected] and [email protected] ( fwhm ) and @xmath103 ( figure [ fig : isimu ] ) . the simulations show two features . first , imfit tends to return major / minor axis ratio of @xmath104 even for input axis ratio of 1 . second , imfit also tends to report smaller sizes for larger input circularized sizes such as [email protected] and [email protected] . the latter is partly because of missing flux in our high - res images and partly because their detection of @xmath105 is too low to fit the extended emission . in light of these simulations , the sizes and ellipticities measured by imfit on the stacked images are not inconsistent with symmetric gaussian emission ( minor / major axis ratio @xmath88 1 ) . given the measured ellipticity for the _ stacked all _ and _ stacked faint _ are 0.7 and 0.5 , respectively , and their @xmath8 are [email protected] and [email protected] , respectively , the simulations indicate that they can in fact be symmetric gaussians . when we see the input - output ellipticity plot for @[email protected] ( bottom in figure [ fig : isimu ] ) , the measured ellipticity for the _ stacked faint _ looks off 1@xmath40 error of the ellipticity of 1 but with in 1.4@xmath40 . next , we investigate individual asxdf sources with @xmath89 10@xmath40 detections in the high - res alma image , asxdf1100.053.1 and 013.1 . asxdf1100.053.1 has a size estimated via its visibility data of [email protected] and shows an ellipticity of 0.5 via imfit . because of our simulations we can not exclude the possibility of symmetric gaussian emission in asxdf1100.053.1 , but an ellipticity of @xmath107 seems to be more plausible . according to our simulations for @[email protected] ( bottom in figure [ fig : isimu ] ) , the measured ellipticity of asxdf1100.053.1 has a probability of only 1.3 per cent that asxdf1100.053.1 has ellipticity of @xmath108 . thus asxdf1100.053.1 has ellipticity of @xmath107 plausibly . moving on to asxdf1100.013.1 , there is not a mock source with the size of asxdf1100.013.1 , [email protected]@xmath920@xmath614 , in the simulation for an input circularized size of [email protected] . this implies that asxdf1100.013.1 may have a starburst region more concentrated in the center than a gaussian profile . these shapes that are unlikely to be symmetric gaussians may be giving us a hint of complex star - forming structure in the small emission area . -amp plot and ` actual ' sizes derived by our monte carlo simulation in noise visibility data for alma sources with 10 and 15@xmath40 alma continuum detections . grey dots mark mock sources with 15@xmath40 detections . this plot shows how the input size for mock sources compares with the size measured by fitting to the @xmath79-amp plot . error bars show 1@xmath40 for the input source size distributions . the plot indicates that measurements for low signal - to - noise sources get less effective at @[email protected] , requiring larger corrections . ( bottom ) relationship between intrinsic minor / major axis ratio and measured circularized size by @xmath79-amp plot based on another simulation . we plot adding offset of @xmath1090.01 , 0 and @xmath1100.01 to @xmath111/@xmath112 for visibility . [ fig : simu ] ] smgs , as measured directly in dust continuum at 1100@xmath36 m , in comparison with the radio sizes @xcite and co emission - line sizes @xcite of @xmath1133 smgs . the sizes measured for @xmath27 smgs are on average about half of those measured in the radio and co for @xmath1133 smgs . in this plot , we have not applied any correction by a possible difference between co line , millimeter and radio continuum emissions which is discussed in section[sec : size ] . [ fig : size ] ] -distance @xmath42 200k@xmath78 expected from the size by @xmath79-amplitude plot and flux measured by imfit on the high - res image . error in flux by visibility comes from the measured size uncertainty shown in figure [ fig : uvamp ] . error in flux by imfit is output by imfit . the dashed gray line shows flux by [email protected]@xmath92flux by visibility . [ fig : comp ] ] 0.01 , 0 and @xmath1100.01 to @xmath111/@xmath112 for visibility . [ fig : isimu ] ] our studies have revealed that the typical physical size ( median , @xmath8 ) of starburst nuclei in @xmath156 smgs is 0.67@xmath9 kpc ( table [ tbl-1 ] ) . in the conversion from intrinsic angular source size to physical scale , we assume uniform redshift probability at @xmath626 for sources without photo-@xmath115 and within @xmath59 error of photo-@xmath115 for the sources with photo-@xmath115 . figure [ fig : size ] reveals that our measured sizes are more than a factor @xmath116 smaller than those of smgs at @xmath1133 ( median radius of [email protected] or 2.5kpc as measured via radio continuum @xcite or @xmath117 or 2kpc as measured via high-@xmath55 co emission these radio and co sizes were measured by gaussian fitting , as with our measurements , so the comparison is fair . errors in calibration cause smearing in interferometric data , but we have no reason to suspect that these larger measurements are due to flawed calibration . if radio and ( sub)millimeter continuum and co line emission trace star - forming regions , figure [ fig : size ] implies that @xmath243 and @xmath27 smgs have different characteristic sizes . however , local galaxies were also reported to have smaller fir sizes than co line(s ) ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and radio continuum ( section 5.1.1 in * ? ? ? note that the fir / radio size ratio of 0.86 in local star forming galaxies shown in @xcite is not sufficient to explain the difference in the millimeter sizes in our sample and the radio sizes in the previous studies . in addition , the sizes of radio - detectable smgs may be affected by radio emission related to radio - loud active galactic nuclei ( e.g. * ? ? ? a kolmogorov - smirnov ( ks ) test gives a probability of 0.3 percent that the differences between radio and millimeter sizes could arise by chance . another ks test with radio size correction due to the empirical fir / radio size ratio of 0.86 gives 3.5 per cent that the differences between radio and millimeter sizes could arise by chance . given a different scale at @xmath243 and @xmath156 , the probability of 3.5 percent is the upper limit , and thus the difference in the size distributions is significant with @xmath118 96.5 percent . about co sizes in figure [ fig : size ] , @xcite presented that the co sizes in @xcite and their radio sizes are consistent by a ks - test providing a probability of 84 percent . recently sub - millimeter continuum ( 870 @xmath36 m ) source size measurements by alma of scuba2 sources including 23 smgs with @xmath11810@xmath40 detections covering @xmath119 ( @xmath22 ; median ) is also reported @xcite . their alma data were taken by an array configuration similar to our schedule block 1 yielding a median synthesized beam of [email protected]@[email protected] with the benefit of shorter observing wavelength than ours . their sample consists mainly of smgs with optical / nir - detections and photo-@xmath115 by optical / nir data contrary to our sample consisting mainly of smgs faint at optical / nir wavelength . their sample is typically twice brighter ( 5.7@xmath9210@xmath75 @xmath120 ; median ) than our sample in infrared luminosity . they report a median size of [email protected]@xmath1010.04 ( major axis ; fwhm ) and [email protected] ( @xmath121 ) by gaussian fitting in images not visibilities . the median in scuba2 sample ( @xmath122 1.2kpc ) seems to be @xmath123 larger than that in our sample ( @xmath124 0.67kpc ) . however , we need to take into account the fact that they measured @xmath121 of major axis and we measured circularized @xmath125 . then we can not compare our sizes with theirs more in details here , but both of our sample and scuba2 sources show smaller fir continuum sizes of star forming region of smgs than the previous radio and co sizes in spite of the different luminosity and redshift distributions in the two samples . in order to reveal the possible relation in fir - continuum size and redshift ( and @xmath126 ) , we need higher - angular resolution imaging of smgs with various properties by alma . @xcite suggested a plausible evolutionary connection between @xmath226 smgs as merger - driven ulirgs and @xmath16 cqgs , based on the following facts : i ) the star - formation history of @xmath16 cqgs @xcite matches @xmath276 smgs ; ii ) the nir sizes of @xmath14 smgs are compact enough for them to be progenitors of cqgs ; iii ) simulations suggest that major mergers at @xmath14 can generate compact stellar components ( e.g. * ? ? ? our results provide direct evidence that starbursts in @xmath27 smgs are compact . we plot the sizes of @xmath14 smgs alongside the nir sizes of compact star - forming galaxies ( csfgs ) at @xmath102.5 @xcite and cqgs @xcite as a function of redshift in figure [ fig : sizered ] . the size of the starburst region in @xmath226 smgs is comparable to ( or smaller than ) the ( nir ) size of the stellar component in @xmath16 cqgs , supporting the idea that compact dust - obscured starbursts in @xmath226 smgs generate the extremely compact , dense stellar components of cqgs . compact star - forming galaxies at @xmath102.5 were reported recently as a possible progenitors of cqgs , based on nir spectroscopy of emission lines and nir source size measurements , but without direct size measurements of the star - forming region . csfgs have similarities with cqgs : csfgs have similar nir structural profiles as cqgs @xcite and the formation redshift of csfgs is also similar to that of cqgs @xcite . these facts could imply that @xmath14 smgs evolve into cqgs via csfgs ( figure [ fig : sizered ] ) . note that @xcite suggests the posibility of disk instability ( e.g. * ? ? ? * ; * ? ? ? * ) as another path to make csfgs and cqgs , based on simulations , in addition to major mergers at @xmath14 . our results make the evolutionary scenario suggested by @xcite more plausible . given that @xmath16 cqgs are thought to evolve into local giant ellipticals mostly via dry mergers ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , our results indicate that local giant ellipticals probably experienced an smg phase at @xmath156 . a discussion based on the surface sfr density ( @xmath127 ) is helpful to understand the origin of the compact but huge star - formation activity in @xmath14 smgs . we derived @xmath127 of our sample using estimated @xmath8 ( table[tbl-1 ] ) . @xmath127 of our sample are in the range of @xmath8930600m@xmath3yr@xmath5kpc@xmath13 with a median of 100@xmath12m@xmath3yr@xmath5kpc@xmath13 . we can find that asxdf sources with a millimeter size of @[email protected] ( fwhm ) show large uncertainty in their @xmath127 ( table[tbl-1 ] ) . these large uncertainties in @xmath127 come from the large fraction of their size errors to their millimeter sizes which contribute to @xmath127 by @xmath128 . first we compare with local galaxies . given that @xmath127 for local merger - driven ( u)lirgs is 54500m@xmath3yr@xmath5kpc@xmath13 with a median of 29@xmath129m@xmath3yr@xmath5kpc@xmath13 and that @xmath127 for local disks is 0.011m@xmath3yr@xmath5kpc@xmath13 with a median of 0.04@xmath130m@xmath3yr@xmath5kpc@xmath13 @xcite from the surface infrared luminosity densities of local galaxies in @xcite for a chabrier imf @xcite . ] , @xmath27 smgs are similar to local ( u)lirgs ( figure [ fig : ssfrd ] ) . a ks test gives a probability of 3.5 percent that the differences between @xmath127 distributions of high-@xmath115 smgs and local ( u)lirgs could arise by chance , and thus the two distribution are consistent with a significant level of 3.5 percent . the range of the infrared luminosities of local ( u)lirg sample is 10@xmath131 @xmath120 which is a little bit fainter than that of our sample . it is worth mentioning that a brighter half of local ( u)lirgs with 10@xmath132 @xmath120 shows more similar @xmath127 distribution to that of high-@xmath115 smgs ; a ks test gives a probability of 17.1 percent that the differences could arise by chance indicating that the @xmath127 distributions of local brighter ( u)lirgs and high-@xmath115 smgs are consistent with a significant level of 17.1 percent . on the other hand , the remaining fainter local lirgs show a less similar @xmath127 to high-@xmath115 smgs ; a ks test shows a probability of 0 percent that the differences could arise by chance . a ks test also shows that @xmath127 distributions of local disks and high-@xmath115 smgs do not match with a probability of 0 indicating that star - formation of high-@xmath115 smgs is different from that in local disks . next we compare with high-@xmath115 extended disk galaxies . we took four bzk galaxies at @xmath1331.5 with co @xmath55=21 sizes from @xcite and 14 high - redshift disk galaxies at @xmath242 with co @xmath55=32 sizes from @xcite as our sample of high - redshift disk galaxies . these co sizes have been derived assuming gaussian profiles . the sample of 18 high-@xmath115 disk galaxies is among 42 disk galaxies with co detections in @xcite including 6 bzks by @xcite . the remaining 24 sources tend to be observed by low angular resolutions and are unresolved in co images . here we adopt the distribution of co sizes of the 18 high-@xmath115 disk galaxies as the representative of the 42 sources because of following facts . @xcite shows that a ratio of @xmath121(optical)/@xmath121(co ) is @xmath134 . we checked that optical size distributions of the high-@xmath115 disk galaxies with co sizes and those without co sizes are consistent with a significant level of 99.7 percent by a ks test . the sfrs of these galaxies are @xmath135500m@xmath3yr@xmath5 and their sizes ( @xmath121 ) range within 212kpc . the distribution of @xmath127 for the high - redshift disk galaxies ranges 0.1 - 7.0m@xmath3yr@xmath5kpc@xmath13 with a median of 0.5@xmath136m@xmath3yr@xmath5kpc@xmath13 . a ks test shows that there is a probability of 0 percent that the differences between @xmath127 distributions of high-@xmath115 smgs and the high-@xmath115 extended disks could arise by chance . we should take into account a possible difference in co and ( sub)millimeter sizes in high-@xmath115 smgs and star - forming galaxies . however , we do not know a size correction factor of co/(sub)millimeter sizes at this moment . then we adopt here co/(sub)millimeter size ratio of 2.9 and 1.7 derived from co sizes of smgs in @xcite and millimeter sizes in our asxdf sources , and submillimeter sizes of smgs in @xcite , respectively . the estimated @xmath127 for the size correction factor of 2.9 is 0.859m@xmath3yr@xmath5kpc@xmath13 with a median of 4m@xmath3yr@xmath5kpc@xmath13 and one for the size correction factor of 1.7 is 0.320m@xmath3yr@xmath5kpc@xmath13 with a median of 1m@xmath3yr@xmath5kpc@xmath13 . then the @xmath14 smgs seem to have typically @xmath13725@xmath92 larger @xmath127 than local and high-@xmath115 disks . ks tests gives probabilities of 0 percent that @xmath127 of high-@xmath115 extended disks with both of the size correction factor of 2.9 and 1.7 are consistent with that of high-@xmath115 smgs . figure[fig : ssfrd ] shows the distributions of @xmath127 of our sample in comparison with local ( u)lirgs , extended disks in high and low redshifts . we derived the expected distribution of @xmath127 of @xmath27 smgs by bootstrapping in order to take into account the large uncertainty of @xmath127 of @xmath27 smgs ( table[tbl-1 ] ) . this large uncertainty comes from the limit of size measurement in our data set . figure[fig : ssfrd ] shows that our sample has @xmath127 similar to local ( u)lirgs rather than those of extended disks at high and low redshifts . even though the compact star - forming region could consist of more compact clumps spreading in @xmath850.31.3kpc as some local ulirgs show ( e.g. * * ; * ? ? ? * ) , almost all of fir continuum in asxdf sources come from one compact region as we presented in section3.2 . then the fact that the @xmath127 of @xmath14 asxdf sources is similar to that of local ( u)lirgs indicates that @xmath14 smgs also harbor a starburst nuclei as local ( u)lirgs do . although local ( u)lirgs are widely thought to be merger - driven , the fact that @xmath14 smgs harbor a compact starburst nuclei as local ( u)lirgs harbor does not imply that the compact starbursts in @xmath14 smgs are triggered by mergers . it is because simulations suggest that disk instability also trigger compact starbursts at high redshift . recently , disk instabilities have been proposed to play important role in triggering starbursts . for example , the galform semi - analytic model of galaxy formation including both types of starbursts merger - driven and disk instability - driven @xcite , and predicts that disk instabilities are the main channel for triggering starbursts at high-@xmath115 that would be seen as smgs @xcite . older versions of the same semi - analytic model argued for merger - driven starbursts as the main formation channel of smgs @xcite . the predicted sizes resulting in starbursts from both disk instabilities and mergers in the galform are in the range @xmath138 kpc ( @xmath121 ) . however , the simplicity of the angular momentum evolution models applied to semi - analytic model prevents us from ruling out the possibility that the high sfr surface densities can only be achieved during galaxy mergers . thus at this stage we can not distinguish the trigger of starbursts of @xmath14 smgs by their compact size . detailed hydro - dynamic simulations addressing this issue are necessary to shed light into how varied are the progenitors of @xmath13336 smgs . we also need further co and continuum observations of these galaxies with higher angular resolution and sensitivity by alma to reveal the trigger of high-@xmath115 smgs . regardless of the triggering mechanism of the @xmath14 smgs , the fact that these are very compact starburst regions supports the evolutionary link between @xmath14 smgs and @xmath16 cqgs proposed by @xcite . ) distribution of @xmath14 smgs in comparison with local ( u)lirgs and disk galaxies @xcite and high-@xmath115 extended disks @xcite . here , we have not applied any correction of a possible size difference between co line and ( sub)millimeter continuum sizes in high-@xmath115 disks . we derived the expected distribution of @xmath127 of @xmath27 smgs by bootstrapping in order to take into account the large uncertainty of @xmath127 of @xmath27 smgs ( table [ tbl-1 ] ) . our @xmath14 smgs show @xmath127 distribution similar to local ( u)lirgs . [ fig : ssfrd ] ] we have exploited new alma 1100-@xmath36 m continuum data to measure the size of dusty , starburst regions in a sample of 13 smgs at @xmath27 . the radii of @xmath226 smgs with @xmath139@xmath140l@xmath3 ranges from [email protected] to [email protected] ( fwhm ) with a median of [email protected]@xmath141 , corresponding to a median circularized effective radius @xmath8 of 0.67@xmath142kpc . our results demonstrate that the star - forming regions of @xmath27 smgs are more than 2@xmath92 smaller than measured previously using radio continuum and co data for @xmath243 smgs . our discovery of compact starbursts in @xmath27 smgs supports the evolutional scenario proposed by @xcite wherein @xmath14 smgs evolve into the most massive ellipticals in the local universe , via dry merging of the cqg population seen at @xmath10 , meaning that we can now trace the evolutionary path of the most massive galaxies over a period encompassing @xmath17% of the age of the universe . this paper makes use of the following alma data : ads / jao.alma#2012.1.00326.s . alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) , in cooperation with the republic of chile . the joint alma observatory is operated by eso , aui / nrao and naoj . soh ikarashi is supported by the alma japan research grant of naoj chile observatory , naoj - alma-0036 . this work was supported by grant - in - aid for jsps fellows number 25 - 10420 . rob ivison acknowledges support from the european research council in the form of advanced investigator programme , cosmicism , 321302 . yoichi tamura is supported by jsps kakenhi no . this work has been partly supported by mexican conacyt research grant cb-2011 - 01 - 16729 . daisuke iono is supported by the 2015 inamori research grants program barro , g. , faber , s. m. , prez - gonzlez , p. g. , et al . 2014a , , 791 , 52 barro , g. , trump , j. r. , koo , d. c. , et al . 2014b , , 795 , 145 baugh , c. m. , lacey , c. g. , frenk , c. s. , et al . 2005 , , 356 , 1191 bezanson , r. , van dokkum , p. g. , tal , t. , et al . 2009 , , 697 , 1290 biggs , a. d. , & ivison , r. j. 2008 , , 385 , 893 blain , a. w. , chapman , s. c. , smail , i. , & ivison , r. 2004 , , 611 , 725 carollo , c. m. , bschorr , t. j. , renzini , a. , et al . 2013 , , 773 , 112 caputi , k. i. , dunlop , j. s. , mclure , r. j. , et al . 2012 , , 750 , ll20 carilli , c. l. , & yun , m. s. 1999 , , 513 , l13 ceverino , d. , dekel , a. , & bournaud , f. 2010 , , 404 , 2151 chabrier , g. 2003 , , 115 , 763 chapman , s. c. , blain , a. w. , smail , i. , & ivison , r. j. 2005 , , 622 , 772 chen , c .- c . , smail , i. , swinbank , a. m. , et al . 2015 , , 799 , 194 cowley , w. i. , lacey , c. g. , baugh , c. m. , & cole , s. 2015 , , 446 , 1784 lagos , c. d. p. , bayet , e. , baugh , c. m. , et al . 2012 , , 426 , 2142 daddi , e. , bournaud , f. , walter , f. , et al . 2010 , , 713 , 686 daddi , e. , renzini , a. , pirzkal , n. , et al . 2005 , , 626 , 680 dekel , a. , sari , r. , & ceverino , d. 2009 , , 703 , 785 elbaz , d. , dickinson , m. , hwang , h. s. , et al . 2011 , , 533 , aa119 gonzalez - perez , v. , lacey , c. g. , baugh , c. m. , et al . 2014 , , 439 , 264 hatsukade , b. , kohno , k. , aretxaga , i. , et al . 2011 , , 411 , 102 hughes , d. h. , aretxaga , i. , chapin , e. l. , et al . 2002 , , 335 , 871 hughes , d. h. , serjeant , s. , dunlop , j. , et al . 1998 , , 394 , 241 ikarashi , s. , kohno , k. , aretxaga , i. , et al . 2013 , new trends in radio astronomy in the alma era : the 30th anniversary of nobeyama radio observatory , 476 , 265 ikarashi , s. , kohno , k. , aguirre , j. e. , et al . 2011 , , 415 , 3081 iono , d. , peck , a. b. , pope , a. , et al . 2006 , , 640 , l1 ivison , r. j. , smail , i. , papadopoulos , p. p. , et al . 2010 , , 404 , 198 ivison , r. j. , greve , t. r. , dunlop , j. s. , et al . 2007 , , 380 , 199 ivison , r. j. , smail , i. , le borgne , j .- f . , 1998 , , 298 , 583 krogager , j .- k . , zirm , a. w. , toft , s. , man , a. , & brammer , g. 2014 , , 797 , 17 lindroos , l. , knudsen , k. k. , vlemmings , w. , conway , j. , & mart - vidal , i. 2015 , , 446 , 3502 nelson , e. , van dokkum , p. , franx , m. , et al . 2014 , , 513 , 394 newman , a. b. , ellis , r. s. , bundy , k. , & treu , t. 2012 , , 746 , 162 oliver , s. j. , bock , j. , altieri , b. , et al . 2012 , , 424 , 1614 onodera , m. , daddi , e. , gobat , r. , et al . 2010 , , 715 , l6 oser , l. , naab , t. , ostriker , j. p. , & johansson , p. h. 2012 , , 744 , 63 riechers , d. a. , bradford , c. m. , clements , d. l. , et al . 2013 , , 496 , 329 rujopakarn , w. , rieke , g. h. , eisenstein , d. j. , & juneau , s. 2011 , , 726 , 93 sakamoto , k. , wang , j. , wiedner , m. c. , et al . 2008 , , 684 , 957 sakamoto , k. , ho , p. t. p. , iono , d. , et al . 2006 , , 636 , 685 sakamoto , k. , scoville , n. z. , yun , m. s. , et al . 1999 , , 514 , 68 simpson , j. m. , smail , i. , swinbank , a. m. , et al . 2015 , , 799 , 81 simpson , j. m. , swinbank , a. m. , smail , i. , et al . 2014 , , 788 , 125 smail , i. , ivison , r. j. , & blain , a. w. , 1997 , , 490 , l5 swinbank , a. m. , simpson , j. m. , smail , i. , et al . 2014 , , 438 , 1267 tacconi , l. j. , neri , r. , genzel , r. , et al . 2013 , , 768 , 74 tacconi , l. j. , genzel , r. , smail , i. , et al . 2008 , , 680 , 246 tacconi , l. j. , neri , r. , chapman , s. c. , et al . 2006 , , 640 , 228 targett , t. a. , dunlop , j. s. , cirasuolo , m. , et al . 2013 , , 432 , 2012 toft , s. , smoli , v. , magnelli , b. , et al . 2014 , , 782 , 68 van dokkum , p. g. , franx , m. , kriek , m. , et al . 2008 , , 677 , l5 wuyts , s. , cox , t. j. , hayward , c. c. , et al . 2010 , , 722 , 1666 wilson , c. d. , rangwala , n. , glenn , j. , et al . 2014 , , 789 , l36 wilson , c. d. , petitpas , g. r. , iono , d. , et al . 2008 , , 178 , 189 younger , j. d. , fazio , g. g. , wilner , d. j. , et al . 2008 , , 688 , 59 c c c c c c c c c c c c name & r.a . & dec . & snr&@xmath143 & sfr@xmath144 & @xmath145 & photo & & @xmath8@xmath146 & @xmath127 + & & & & & & & @xmath115 & raw & corrected & & + & ( j2000 ) & ( j2000 ) & & ( mjy ) & ( m@xmath3yr@xmath5 ) & ( @xmath147l@xmath3 ) & & ( arcsec ) & ( arcsec ) & ( kpc ) & ( m@xmath3yr@xmath5kpc@xmath37 ) + + asxdf1100.013.1 & 02:16:45.86 & @xmath1095:03:47.2 & 18.5 & [email protected] & 440@xmath148 & 4.4@xmath149 & @xmath1184.8 & 0.298 & 0.32@xmath150 & 1.1@xmath151 & 60@xmath152 + asxdf1100.027.1 & 02:17:20.95 & @xmath1095:08:37.2 & 14.8 & [email protected] & 400@xmath153 & 4.0@xmath154 & 2.80@xmath155 & 0.076 & 0.10@xmath156 & 0.4@xmath149 & @xmath157 + asxdf1100.045.1 & 02:18:16.04 & @xmath1094:54:02.8 & 13.0 & [email protected] & 360@xmath153 & 3.6@xmath154 & @xmath1185.5 & @xmath870.039@xmath158 & 0.10@xmath156 & 0.3@xmath159 & 610@xmath160 + asxdf1100.045.2 & 02:18:14.89 & @xmath1094:54:03.9 & 12.9 & [email protected] & 330@xmath153 & 3.3@xmath159 & @xmath1183.4 & 0.307 & 0.31@xmath156 & 1.2@xmath154 & 39@xmath161 + asxdf1100.049.1 & 02:17:32.86 & @xmath1094:57:00.8 & 12.3 & [email protected] & 320@xmath162 & 3.2@xmath159 & @xmath1183.8 & 0.243 & 0.28@xmath163 & 0.9@xmath164 & 61@xmath165 + asxdf1100.053.1 & 02:16:48.20 & @xmath1094:58:59.6 & 27.6 & [email protected] & 610@xmath166 & 6.1@xmath167 & @xmath1184.4 & 0.300 & 0.28@xmath168 & 0.9@xmath169 & 91@xmath170 + & 25.0 & [email protected] & 340@xmath162 & 3.4@xmath159 & & 0.170 & 0.18@xmath171 & 0.6@xmath172 & 160@xmath173 + & 31.0 & [email protected] & 420@xmath174 & 4.2@xmath159 & & 0.240 & 0.24@xmath175 & 0.8@xmath151 & 110@xmath176 + + asxdf1100.073.1 & 02:18:10.04 & @xmath1095:11:31.7 & 10.4 & [email protected] & 220@xmath177 & 2.2@xmath151 & @xmath1183.6 & 0.308 & 0.34@xmath178 & 1.2@xmath179 & 26@xmath180 + asxdf1100.083.1 & 02:17:12.42 & @xmath1095:03:59.4 & 13.7 & [email protected] & 360@xmath162 & 3.6@xmath159 & @xmath1184.1 & 0.355 & 0.38@xmath181 & 1.3@xmath154 & 34@xmath182 + asxdf1100.090.1 & 02:17:23.04 & @xmath1094:57:29.9 & 11.3 & [email protected] & 290@xmath162 & 2.9@xmath159 & @xmath1183.2 & @xmath870.039@xmath158 & 0.10@xmath183 & 0.4@xmath149 & 360@xmath184 + asxdf1100.110.1 & 02:17:43.59 & @xmath1095:04:10.3 & 12.7 & [email protected] & 275@xmath148 & 2.8@xmath149 & 4.98@xmath185 & @xmath870.039@xmath158 & 0.10@xmath156 & 0.3@xmath149 & 420@xmath186 + asxdf1100.127.1 & 02:17:33.36 & @xmath1094:48:43.8 & 13.3 & [email protected] & 320@xmath162 & 3.2@xmath159 & @xmath1183.8 & 0.192 & 0.22@xmath187 & 0.7@xmath164 & 99@xmath188 + asxdf1100.230.1 & 02:17:59.39 & @xmath1094:45:53.1 & 11.3 & [email protected] & 350@xmath189 & 3.5@xmath151 & 3.50@xmath190 & @xmath870.039@xmath158 & 0.10@xmath183 & 0.4@xmath149 & 380@xmath191 + asxdf1100.231.1 & 02:17:59.65 & @xmath1094:46:49.7 & 26.8 & [email protected] & 400@xmath162 & 4.0@xmath159 & @xmath1185.9 & 0.107 & 0.12@xmath192 & 0.4@xmath172 & 450@xmath193 +	we report the source size distribution , as measured by alma millimetric continuum imaging , of a sample of 13 aztec - selected submillimeter galaxies ( smgs ) at @xmath06 . their infrared luminosities and star - formation rates ( sfr ) are @xmath12@xmath2l@xmath3 and @xmath4600m@xmath3 yr@xmath5 , respectively . the sizes of these smgs range from [email protected] to [email protected] , with a median of [email protected]@xmath7 ( fwhm ) , corresponding to a median circularized effective radius ( @xmath8 ) of 0.67@xmath9kpc , comparable to the typical size of the stellar component measured in compact quiescent galaxies at @xmath10 ( cqgs ) @xmath11kpc . the median surface sfr density of our smgs is 100@xmath12m@xmath3yr@xmath5kpc@xmath13 , comparable to that seen in local merger - driven ( u)lirgs rather than in extended disk galaxies at low and high redshifts . the discovery of compact starbursts in @xmath14 smgs strongly supports a massive galaxy formation scenario wherein @xmath156 smgs evolve into the compact stellar components of @xmath16 cqgs . these cqgs are then thought to evolve into the most massive ellipticals in the local universe , mostly via dry mergers . our results thus suggest that @xmath14 smgs are the likely progenitors of massive local ellipticals , via cqgs , meaning that we can now trace the evolutionary path of the most massive galaxies over a period encompassing @xmath17% of the age of the universe .
the focus of this paper is three - fold : 1 . to review an historically significant problem in analytic number theory , 2 . to describe a particular branch of @xmath2 , 3 . using ( 2 ) , solve ( 1 ) in a more elementary way . in 1859 , bernhard riemann published `` on the number of primes less than a given magnitude''@xmath4 , and introduced an analytic continuation of the euler product he called the zeta function : @xmath5 . using @xmath5 , infinite product representations of the @xmath6 and @xmath7 functions and relying on advanced principles of analysis , riemann went on to derive an analytic expression for a function he called @xmath0 equal to the number of prime numbers under @xmath1 . however , if a branch of @xmath2 is carefully chosen , riemann s expression for @xmath0 can be derived via the residue theorem using simpler techniques of analysis . this approach is described below . throughout this paper , a complex variable is represented as @xmath8 , and the non - trivial zeros of the zeta function with positive imaginary component as @xmath9 ( or just @xmath10 in some cases ) , such that @xmath11 . the principal branch of the complex logarithm is denoted by log and other branches are given by log . common in complex analysis , log represents logarithm to base e , that is log(e)=1 . that convention is used here . in all cases , the variable x is taken to be a real number greater than one . riemann derives his prime counting function by beginning with the euler product : @xmath12 taking logarithms of the above gives @xmath13 and remembering @xmath14 for @xmath15z@xmath161 , we obtain @xmath17 replacing each prime power with @xmath18 we can express equation as @xmath19 with @xmath20 @xmath0 is equal to the number of primes strictly less than @xmath1 and as is common in fourier theory , at points of discontinuity , namely at prime numbers , it is defined to be one - half the difference between it s old and new value . with this definition of @xmath0 , we see @xmath21 begins at 0 for x=0 and increases by a jump of 1 at primes , by a jump of 1/2 at prime squares @xmath22 , by a jump of @xmath23 at prime cubes @xmath24 , etc . that is , @xmath21 is zero for @xmath25 , is @xmath26 for @xmath27 , is 1 for @xmath28 , is @xmath29 for @xmath30 , is @xmath31 for @xmath32 , is @xmath33 for @xmath34 , is @xmath35 for @xmath36 , is @xmath37 for @xmath38 and so on . from @xmath21 , one can find @xmath0 via mbius inversion to obtain @xmath39 with @xmath40 a positive integer such that @xmath41 . the function f(x ) is riemann s prime counting function and riemann s derivation of an analytic expression for @xmath21 and thus @xmath0 , is the principal result of his paper . he accomplished this by inverting via fourier inversion giving @xmath42 in order to evaluate , riemann expressed @xmath5 in terms of the @xmath43 function : @xmath44 and solved for @xmath45 . he then expressed both @xmath7 and @xmath46 in terms of their infinite product representations and proceeded to evaluate the resulting integrals directly to obtained . @xmath47 is the corrected version . ] @xmath48+\int_x^{\infty } \frac{1}{t\left(t^2 - 1\right ) \text{log}(t ) } \ dt-\text{log}\,(2 ) \label{eqn003}\ ] ] in which @xmath49 is the logarithmic integral and @xmath50 with @xmath51 the exponential expressions are obtained by first considering @xmath52 where the contour @xmath53 is along the real axis from zero to @xmath1 with an indentation above the ( simple ) pole at @xmath54 for @xmath55 giving rise to the @xmath56 term , or below it for @xmath57 and giving rise to the @xmath58 term . letting @xmath59 gives @xmath60 where the path of integration for the integral on the right is now along the real axis with an indentation above or below the origin . making the substitution @xmath61 , then @xmath62 in which @xmath63 the center integral of the last expression tends to zero leaving the desired result . for a further explanation of , see edwards and ingham @xcite . although riemann s work represents a beautiful case study in analysis , his approach to evaluating was based on advanced methods of analysis . in this paper , is evaluated more simply using a holomorphic branch of @xmath2 . the geometry of this branch is investigated and the results used to derive an integral expression for @xmath21 which is then evaluated via the residue theorem . the function @xmath64 is holomorphic in the half - plane @xmath65 since @xmath5 is non - zero there . however , there exists a holomorphic branch that extends to the left half - plane . proof of the following theorem can be found in @xcite . if @xmath66 is a nowhere vanishing holomorphic function in a simply connected region @xmath67 , then there exists a holomorphic function @xmath68 on @xmath67 such that @xmath69 the function @xmath70 is given by @xmath71 , and determines a `` branch '' of that logarithm . also : @xmath72 [ theorem001 ] the trivial zeros of @xmath5 are located along the negative real axis at the points @xmath73 . the non - trivial zeros are symmetric in the critical strip @xmath74 to the line @xmath75 , and @xmath76 has one simple pole at @xmath54 . therefore , we can define @xmath67 to be the complex plane with branch cuts at the real axis extending from @xmath77 to one , and branch cuts of the form @xmath78 for each zero on the critical line and branch cuts between zeros for those off the critical line . we then construct a contour @xmath79 around @xmath67 . this contour is shown in figure ( [ fig : file1 ] ) and explained below . then @xmath80 is holomorphic in @xmath67 with @xmath81 where the set @xmath82 represents the branch cuts , and the function @xmath83 is now a continuous and analytic function of @xmath84 throughout @xmath85 . therefore , @xmath86 with the path of integration remaining in @xmath67 . in order use @xmath2 in this analysis , we must first understand the geometry of @xmath87 which we do now . @xmath88 is continuous except at the zeros of the zeta function and the point @xmath54 . the imaginary sheet , @xmath87 , is however more complicated , and a description of this sheet in the neighborhood of the branch cuts is needed in order to use the residue theorem to derive riemann s expression for @xmath21 . @xmath87 along a section of the black contours of @xmath79 is shown in figure ( [ fig : realaxiscloseup ] ) . note the stair - step geometry . the dark black trace is the contour below the real axis and the gray trace , above the real axis . empirical results suggest the difference in argument changes by @xmath89 at each step . we can prove this difference using the following theorem @xcite : let @xmath90 be analytic with a simple pole at @xmath91 and @xmath92 be an arc of a circle of radius @xmath93 and angle @xmath94 centered at @xmath91 . then : @xmath95 [ theorem005 ] consider now @xmath96 and the points @xmath97 and @xmath98 in figure ( [ fig : closeupat2 ] ) where @xmath99 and @xmath100 with @xmath101 ( note the indentation around the origin applies only to @xmath102 ) . we then have in the limit as the black contours approach the real axis , along black contours of b(m ) ] @xmath103 now , the first and third integrals are analytic outside some deleted neighborhood @xmath104 of @xmath54 and therefore the integrals in this region cancel as @xmath105 . inside @xmath104 , @xmath106 with @xmath107 for @xmath108 . clearly , the residue of @xmath109 at @xmath54 is @xmath110 . ] now , because of the symmetric form of the limit , i.e. , the terms @xmath111 , the angle @xmath94 of @xmath92 will always extend from @xmath112 to @xmath113 even in the limit as @xmath105 . therefore @xmath114 consider now the remaining two integrals inside @xmath104 . substituting into the integrals , the @xmath115 terms will again because they are analytic , cancel as @xmath105 leaving @xmath116 thus as the black contours in the range @xmath117 approach the real axis , @xmath118 and therefore @xmath119 now , the real part of @xmath5 in the range @xmath117 is negative and the imaginary part of @xmath5 is negative in the range @xmath120 along the upper black contour ( gray contour in figure [ fig : realaxiscloseup ] ) . therefore @xmath87 along the upper black contour in the range @xmath101 must approach @xmath121 since this branch of @xmath83 is continuous and therefore by the expression above , @xmath87 along the lower contour is @xmath122 over the same range . now , in the vicinity of each trivial zero , @xmath123 with the residue of this expression clearly being one . referring to figure ( [ fig : closeupat2 ] ) , consider the points @xmath124 and @xmath125 . then @xmath126 and again because of the symmetric form of the limit , @xmath127 leaving for the integrals : @xmath128 substituting into , we obtain after noting the analytic parts cancel : @xmath129 since @xmath130 . then @xmath131 and @xmath132 therefore the imaginary component of @xmath2 along the upper black contour as it approaches the real axis in the range @xmath133 has the limit zero since we have shown above that @xmath134 for @xmath101 along the upper black contour has a limit of @xmath121 . this same approach can then be applied to the remaining components of the black contour to arrive at the geometry in figure ( [ fig : realaxiscloseup ] ) . @xmath87 in a region around the branch cut at @xmath135 is shown in figure ( [ fig110:subfig : a ] ) . empirical results indicate the red contour ( over the imaginary sheet of @xmath2 ) is @xmath136 that of the blue contour . a similar argument to that in section [ section31 ] can be used to explain this geometry : for a point @xmath137 on the blue contour and @xmath138 on the red contour , @xmath139 and because of symmetry , @xmath140 for a deleted neighborhood surrounding each trivial zero on the critical line , called a `` common zero '' in this paper , we have @xmath141 making the same type of substitution as above gives @xmath142 this leaves @xmath143 and thus @xmath144 which is the value observed empirically . we must also consider the possibility of zeros off the critical line . these are called `` rogue zeros '' in this paper . how would these effect the branch cuts ? we can artificially introduce such zeros by considering the function @xmath145 and numerically solve ( [ inteqn1 ] ) using this function . the plot in figure ( [ fig110:subfig : b ] ) shows @xmath146 $ ] at this branch cut with @xmath147 and @xmath148 . numerical results suggest the difference in argument from @xmath149 to the first zero is @xmath150 and between the zeros is @xmath89 . this same type of geometry would occur with actual rogue zeros and using the same argument above , this @xmath150 and @xmath89 difference is easily explained . we now consider @xmath151 where @xmath79 includes the possibility of zeros off the critical line with the horizontal gray legs of the contour passing over certain lines with imaginary components equal to @xmath152 . since @xmath79 is closed over a function analytic throughout the region of integration , @xmath153 and therefore @xmath154 now @xmath155 over the purple contour and therefore we can replace @xmath2 with @xmath64 in that integral . multiplying by @xmath156 , and taking the limit as @xmath157 , @xmath158 where in , the appropriate limits are taken for each integral . we now go on to evaluate the various color - coded components of this expression . the green contours encircle each non - trivial zero of the zeta function . we allow the diameter of these contours to approach zero and take the following limit : @xmath159 where @xmath160 represents the set of non - trivial zeros of the zeta function . writing this in terms of s=@xmath161 , we have for the common zeros : @xmath162 and @xmath163 for rogue zeros with @xmath164 . the @xmath165 term is bounded by @xmath166 and note as @xmath167 , the denominator tends to @xmath168 with the numerator tending to @xmath169 with @xmath170 being an upper bound on the remaining terms . the zeta function has simple zeros and therefore @xmath171 which tends to zero . this gives @xmath172 independent of the placement of non - trivial zeros . the cyan contours encircle the trivial zeros and yield the following integrals : @xmath173 similar to the integrals over the green indentations , we express these integrals in terms of s=@xmath174 and since these zeros are simple , we obtain a similar limiting process : @xmath175 and therefore @xmath176 the integral around the orange contour at s=1 yields @xmath177 similar to the integrals over the green indentations , we express the integral in terms of s=@xmath178 and arrive at a similar limiting process : @xmath179 the pole at s=1 is a simple pole so that this limit approaches the form @xmath180 therefore @xmath181 we wish to consider the following limit : @xmath182 referring to figure ( [ fig : file1 ] ) , we can write this in terms of two integrals : @xmath183 note each integral traverses surfaces separated by a branch cut with @xmath184 as described in section [ section31 ] . as discussed in that section , @xmath185 over the upper contour is @xmath121 and over the lower contour is @xmath122 as shown in figure ( [ fig : realaxiscloseup ] ) . we then have for the upper contour : @xmath186 with @xmath187 . as @xmath188 , the first term of the last integral tends to @xmath189 and the imaginary component of @xmath190(s ) over the contour near zero tends to @xmath191 . that is , @xmath192 using these values we have @xmath193 in the same way , @xmath87 over the lower contour approaches @xmath194 according to section [ section31 ] , and we obtain for that result : @xmath195 therefore the sum of the integrals over the central brown contours as the diameters tend to zero is @xmath196 we consider two cases : case i takes the zeros on the critical line and case ii considers zeros off the critical line . in the case of zeros on the critical line , the value of log @xmath197(s ) on the blue contour is @xmath198 and the value on the red contour is @xmath199 as discussed in section [ subsection32 ] . in the limit as @xmath200 , the integrals over these contours as @xmath79 extends to infinity are @xmath201 letting @xmath202 , then @xmath203 and with @xmath204 gives @xmath205.\ ] ] for the conjugate contours at @xmath206 , @xmath207\ ] ] with the last integrals expressed in terms of the logarithmic integral as per . in this case we have sets of zeros of the form @xmath208 the presence of these zeros will cause the values between the red and blue contours to differ by @xmath150 between the range @xmath77 to @xmath209 and by @xmath210 between the range @xmath211 as was discussed in section [ subsection32 ] . we obtain for each set of zeros @xmath212 and with @xmath213 , @xmath214 this then becomes @xmath215 with a similar case for the pair with negative imaginary parts . these expressions are again equal to @xmath216 . that is , @xmath217 the set of conjugate zeros gives a similar expression in terms of @xmath218 . therefore we can write for the set of common and rogue zeros , @xmath219 so that for the set of all non - trivial zeros @xmath220 : @xmath221.\ ] ] the integrals over the black contours are determined by the change in argument over these contours described in section [ section31 ] . those over the interval ( -6,-4 ) are @xmath222 , \end{aligned}\ ] ] with @xmath223 being the incomplete gamma function . over the interval @xmath224 , these integrals are @xmath225 . \end{aligned}\ ] ] and for the interval @xmath226 , @xmath227 . \end{aligned}\ ] ] and in general , @xmath228 . \end{aligned}\ ] ] adding the first few terms of this set gives @xmath229 summing all the terms , we have for the black contours in the interval @xmath230 : @xmath231,\ ] ] with the sum converging due to the exponential decay of the gamma integrals . as discussed in section [ section31 ] , in the range ( -2,1 ) , the imaginary surface over the lower contour is @xmath122 and over the upper contour , @xmath121 . therefore , log @xmath197(s ) differs by 2@xmath194i over these two contours . however , the integrand is singular at the origin and thus we take the principal value for these integrals . we then have @xmath232 we wish to prove the following : @xmath233 where the gray boundary is incremented through certain lines crossing the critical strip of the zeta function . proofs of the following theorems can be found in ingham @xcite . there exist a sequence of numbers , @xmath234 , such that : @xmath235 [ thm001 ] for some positive constant @xmath236 . these are the lines at @xmath152 of figure ( [ fig : file1 ] ) . in the region obtained by removing from the half - plane @xmath237 , the interiors of a set of circles surrounding each trivial zero with radius @xmath26 , we have : @xmath238 [ thm002 ] and for @xmath239 : @xmath240 theorem @xmath241 implies that between each integer @xmath242 and @xmath243 , there exists a horizontal contour passing between the zeta zeros over which the order of @xmath109 does not exceed @xmath244 . theorem @xmath245 states this order does not exceed @xmath246 when @xmath247 . contour ( for @xmath248 sufficiently large ) ] figure([fig : tmcontour ] ) shows the top gray contour , @xmath249 , over a finite path of length @xmath250 . then @xmath251 where @xmath252 . that is , for @xmath253 sufficiently negative and @xmath248 sufficiently large , we have the inequality @xmath254 the value of @xmath2 over this contour given by is @xmath255 and therefore @xmath256 where @xmath236 is a positive constant and @xmath257 for @xmath258 . substituting into , we can write for the first integral : @xmath259 since @xmath260 . now consider the second integral and the expression @xmath261 therefore the bound on @xmath2 over this segment becomes @xmath262 by . now , for @xmath263 we have @xmath264 , then @xmath265 and therefore @xmath266 this is because for @xmath267 , @xmath268 . combining and gives @xmath269 and as @xmath270 , the quantity on the left side tends to zero . the same argument then applies to the lower contour at @xmath271 . we wish to first place upper bounds on the quantity @xmath2 over the contours @xmath272 shown at the left in figure ( [ fig : file1 ] ) . for @xmath273 we can write @xmath274 with @xmath275 . using then : @xmath276 for the contour @xmath277 over the vertical interval @xmath278 we have @xmath279 now , @xmath273 is separated from @xmath277 by a branch cut with @xmath280 we can then write @xmath281 and using we obtain @xmath282 and in general : @xmath283 now denote the number of non - trivial zeros of @xmath5 in the range @xmath284 by @xmath285 . a proof of the following theorem , first proposed by riemann , can be found in ingham @xcite : when @xmath286 , @xmath287 [ zerolimit ] thus in the limit as @xmath270 , the number of zeros in the range @xmath288 and thus the number of contours @xmath272 does not have an order which exceeds some constant times the factor of @xmath289 . writing as @xmath290 we have @xmath291 since : @xmath292 and @xmath293 this gives : @xmath294 and as @xmath270 , this limit approaches zero . summing over all the contours @xmath272 as the boundary extends to infinity we obtain : @xmath295 a similar argument then applies to the lower set of vertical contours . we now have all the contours of b(m ) and can write as @xmath296+\sum_{n=2}^{\infty } \gamma [ 0,2n \log ( x)]-\text{log } ( 2 ) . \label{eqn62}\ ] ] riemann obtained @xmath297+\int_x^{\infty } \frac{1}{t\left(t^2 - 1\right ) \text{log}(t ) } \ , dt-\text{log}(2 ) . \label{eqn003b}\ ] ] therefore , if we can show : @xmath298,\ ] ] then the results calculated via the residue theorem are equal to that obtained by riemann . note first if we let @xmath299 in the integral @xmath300 we obtain @xmath301 this leaves us to verify : @xmath302-\int_0^{1\left / x^2\right . } \frac{1}{{\text{log}\,}u } \ , du.\ ] ] now @xmath303 and @xmath304 we therefore wish to show @xmath305 we can write @xmath306 we are left then with showing @xmath307 which we can show by making two change of variables . the first let @xmath308 to obtain @xmath309 now let v =- r : @xmath310 which was to be shown . using a particular holomorphic branch of @xmath2 and the residue theorem , we obtain the same expression for @xmath21 as riemann .	in his paper `` on the number of primes less than a given magnitude''@xcite , bernhard riemann introduced a prime counting function @xmath0 which counts the number of primes under @xmath1 . riemann obtained an analytic expression for @xmath0 by evaluating an inverse laplace transform . his method involved advanced techniques of analysis . however , this transform can be evaluated using the residue theorem when an appropriate branch of @xmath2 is defined . in this paper , a method for constructing a holomorphic branch of @xmath2 extending to the left half - plane is described along with it s geometry surrounding the logarithmic branch points . using this information , an integral representation of @xmath0 is formulated in terms of this branch of @xmath3 which is then evaluated . the results are shown equal to riemann s expression .
my talk is divided into five parts . in the first part the relevance of the chiral fermion issue to fundamental particle physics and to numerical qcd will be explained . in the second part the basic ideas of the overlap construction will be reviewed . in the third part i shall focus on the algebraic ( as opposed to kinematic ) meaning of the overlap . in the fourth part i shall discuss in greater detail the overlap dirac operator on the lattice . in the last part i shall deal with anomalies , the associated berry phases and f " orster - nielsen - ninomiya gauge averaging in the context of naturalness . my main partner in the overlap construction was r. narayanan . i have also collaborated with p. huet , y. kikukawa , a. yamada and p. vranas . important contributions to the overlap development were also made by s. randjbar - daemi and j. strathdee . fermion chirality is one of the most fundamental properties of nature where it appears in association with gauge interactions . this combination can decouple from a more fundamental theory if anomalies cancel . it is however not trivial to achieve this decoupling outside perturbation theory , and for a while it was even thought an impossible feat . the overlap provides an example where the decoupling works in a non - perturbative framework . in a unitary framework the mechanism requires an infinite number of fermion degrees of freedom per unit four space - time volume . it would be a major achievement for lattice field theory if it turned out that a similar mechanism is operative in nature . more details regarding general relevance can be found in @xcite . the overlap is an outgrowth of a long sequence of papers starting from the fundamental discoveries in @xcite , @xcite , @xcite and more specifically @xcite , @xcite , @xcite , @xcite . my first papers on the subject were all written with narayanan who stayed my collaborator for a long time @xcite , @xcite . in parallel , important contributions to the overlap were made in @xcite . later , i continued to simplify the overlap in the vector - like context @xcite . let us start with a formally vector - like gauge theory : @xmath0 @xmath1 and @xmath2 are dirac fermions and the mass matrix @xmath3 is infinite . it has a single zero mode but its adjoint has no zero modes . as long as @xmath4 this setup is stable under small deformations of the mass matrix implying that radiative corrections will not wash the zero mode away . kaplan s domain wall suggests the following realization : @xmath5 where @xmath6 and @xmath7 is fixed at @xmath8 for negative @xmath9 and at @xmath10 for positive @xmath9 ( @xmath11 . there is no mathematical difficulty associated with the discontinuity at @xmath12 . the infinite path integral over the fermions is easily `` done '' : on the positive and negative segments of the real line respectively one has propagation with an @xmath9-independent `` hamiltonian '' . the infinite extent means that at @xmath12 the path integrals produce the overlap ( inner product ) between the two ground states of the many fermion systems corresponding to each side of the origin in @xmath9 . the infinite extent also means infinite exponents linearly proportional to the respective energies - these factors are subtracted . one is left with the overlap formula which expresses the chiral determinant as @xmath13 . the states are in second quantized formalism . by convention , they are normalized , but their phases are left arbitrary . this ambiguity is essential , as we shall see later on . it has no effect in the vector - like case . in first quantized formalism the overlap is : @xmath14 the elements of the matrix @xmath15 are the overlaps between single body wave - functions , @xmath16 . the @xmath17 s span the negative energy subspace of @xmath18 and the @xmath19 s span the negative energy subspace of @xmath20 . i used continuum like notation to emphasize that the hamiltonians are arbitrary regularizations of massive four dimensional dirac operators with large masses of opposite signs . the hamiltonians only enter as defining the dirac seas and there is no distinction between the different levels within each sea ; all that matters is whether a certain single particle state has negative or positive energy . thus , all the required information is also contained in the operators @xmath21 and @xmath22 where @xmath23 is the sign function . thus the @xmath17 s are all the @xmath24 eigenstates of @xmath25 and the @xmath19 s are all the @xmath24 eigenstates of @xmath26 . to switch chiralities one only has to switch the sign of the hamiltonians . this is a result of charge conjugation combined with a particle - hole transformation . when @xmath27 is taken to infinity in lattice units one is left with @xmath28 with no gauge field dependence . on the other hand , @xmath26 always maintains a dependence on the gauge background and its trace gives the topological charge of the gauge field . more recently , a construction starting from the gw relation @xcite has been presented as new @xcite . this is misleading , since the `` new '' construction merely reproduces the structure of the overlap . the only differences are in technicalities surrounding the phase choice of the overlap . these technicalities are not trivial , but i would be surprised if any new particle physics were produced by this effort . to be sure , a successful completion of the phase choice program would be news in mathematical physics , and would make me happy . the proposal of the `` new '' construction was preceded by the observation that the overlap satisfies the gw relation @xcite , @xcite and that the overlap - gw relation is essentially one to one @xcite , @xcite . thus , it should have been obvious that starting from the gw relation will lead to the overlap structure in the chiral case . the insistence of the proponents of the `` new '' construction to obscure its structural equivalence to the overlap has generated much confusion and waste of time , deflecting energy from some valid physics problems that are still open . for example , it is an important open question whether a genuinely different natural construction of chiral gauge theories is possible . the gw relation is a complicated way to write down what is best described as the algebraic content of the overlap . it is much simpler to rephrase and slightly generalize : the setup is that of a `` kato pair '' @xcite , an arrangement further developed in the quantum hall effect context by avron , seiler and simon @xcite and others @xcite . the core objects are the two hermitian reflections , @xmath26 and @xmath25 introduced above . they are equivalent to two orthogonal projectors @xmath29 , the kato pair ( @xmath30 ) . defining h=12(+^ ) , s=12(-^ ) , we arrive at the following fundamental relations : h^2 + s^2 = 1 , \{h , s } = 0 . these relations allow us to write any element in the algebra generated by @xmath31 and @xmath9 as @xmath32 , where @xmath33 is a central element . geometrically , @xmath34 and @xmath35 generate two orthogonal decompositions of the entire space @xmath36 . a natural operator mapping one subspace of one decomposition to another subspace of the other decomposition was introduced by kato . it coincides with overlap dirac operator . d_o = 12 ( 1+^)=^h = h . that it connects the two decompositions is evident from @xmath37 . in the overlap context @xmath38 was found via the following easily proven identity : det d_0 = \|det m\|^2 . the following decomposition of the hilbert space into subspaces invariant under @xmath31 and @xmath9 holds : = ker ( h ) ker ( s ) _ . @xmath39 by virtue of @xmath40 . one also has ker(h)=ker(s-1)ker ( s+1 ) , ker(s)=ker(h-1)ker ( h+1 ) . a special property of the kato pair in the overlap context is that @xmath41 are both gauge field dependent and that for all backgrounds @xmath42 . one can introduce now a definition of index ( ^ , ) = dim ker ( s-1 ) - dim ker ( s+1)= tr s^2n+1 for any integer @xmath43 . to prove @xmath44-independence of the right hand side use @xmath45 and @xmath46 . reversing the roles of @xmath31 and @xmath9 switches the sign of the index . it turns out that the index is exactly the topological charge of the gauge field as sensed by the overlap @xcite . moreover , if @xmath47 and @xmath48 are isolated points of the spectrum of @xmath9 and both @xmath49 and @xmath50 are finite dimensional the definition extends to @xmath51 . until now the roles of @xmath31 and @xmath9 were symmetrical , and there is no massless fermion in sight . the entire setup is pure mathematics . the kinematic part of the construction is in the relationship between the reflections ( or projectors ) and the massive dirac operator . this is the basis of the lattice realization of the overlap idea to be covered in the next section . here , we shall devote a few lines to explain why the kato pair is an algebraic set - up which is essentially equivalent to the ginsparg wilson relation . the ginsparg wilson relation is : \{d , _ 5 } = 2 d_5 r d . one adds to the above the following requirements : ( 1 ) @xmath52-hermiticity , @xmath53 ; ( 2 ) @xmath54=0 $ ] ; ( 3 ) @xmath55 ; @xmath56 is local in the site index . it goes without saying that @xmath57 and @xmath56 transform covariantly under gauge transformations and that @xmath57 should be local in some sense . choosing @xmath58 it is obvious that @xmath38 satisfies the gw relation and the extra requirements with @xmath59 . @xmath26 , being associated with very massive fermions , will be essentially local . in the lattice context , as we shall see in the next section , some gauge backgrounds need to be excluded if locality is to hold absolutely . what may be less obvious is the other direction of the equivalence , namely , that the gw relation implies a kato pair . starting from the gw relation takes one back to the overlap situation at the algebraic level , i.e. , before the explicit form of the reflection @xmath26 is given . the reflection @xmath25 , which is quite free in the overlap , is restricted in the gw case to @xmath52 . the first step is to eliminate @xmath56 , by showing that any solution with @xmath59 produces a solution with an arbitrary @xmath56 . this can be done in at least two ways @xcite . the first starts from the main observation that the basic gw relation @xcite can be written as \{d^-1(r ) - r,_5 } = 0 and therefore setting @xmath60 ensures that @xmath61 is a solution if @xmath62 is . although the inverse of @xmath63 is non - local by the well known no - go theorems , @xmath64 has only zeros at the `` doubler '' locations in the free case , and @xmath56 removes these zeros from @xmath65 , so that @xmath61 ends up local . once @xmath61 is local and doubler free in the free case , locality in the presence of lattice backgrounds that are sufficiently close to the continuum is assured . the second way is to set @xmath66 . so , nothing is lost or gained in principle by fixing @xmath59 . for @xmath59 it is trivial to go back to the reflections : one just looks at the formula for @xmath38 to define the operator @xmath67 $ ] . it is trivial that this operator is hermitian and squares to unity by the basic gw formula @xcite . so , @xmath62 is just another @xmath38 . now it is trivial to extract @xmath31 and @xmath9 , since @xmath25 is given as @xmath52 . also , it is a trivial matter to go to the matrix @xmath15 and the chiral case . the massless fermions enter as a representation of the difference between dirac operators with masses of opposite sign . there are no particular difficulties to regularize _ massive _ dirac fermions , in the continuum or on the lattice . on the lattice the spectra of the dirac operators get compactified and the regularization is fully non - perturbative . first , some notation : let @xmath68 denote @xmath69 link matrices on a finite @xmath70-dimensional hypercubic lattice . the directional parallel transporters @xmath71 are : t _ ( ) ( x ) = u_(x ) ( x+ ) . let @xmath72 with @xmath73 describe a gauge transformation . g(g ) t_(u ) g^(g ) = t_(u^g ) , where , u^g_(x ) = g(x ) u_(x ) g^(x+ ) . a lattice replacement of the massive continuum dirac operator , @xmath74 , is an element in the algebra generated by @xmath75 . thus , @xmath74 is gauge covariant . the wilson dirac operator , @xmath76 can be written as : d_w = m+_(1- v _ ) ; v_^v_=1 ; v_= 1-_2 t_+1+_2 t_^. the hermitian wilson dirac operator is @xmath77 . in terms of the continuum covariant derivative @xmath78 , for lattice spacing @xmath79 , @xmath80 , with no sum on @xmath81 . we choose the following norm definition for matrices @xmath82 : @xmath83 . the norm of a gauge covariant matrix is gauge invariant . @xmath84 is a maximal eigenvalue . our two reflections are defined by : ^= _ m+ ( h_w ( m ) ) = _ d+1 ; =( h_w ( m ) ) with -2<m<0 . for @xmath26 to be defined everywhere we need @xmath85 . this can be assured if the pure gauge action enforces an upper bound on the norms of @xmath86 $ ] @xcite . the following inequality , when meaningful , specifies what constraint would be sufficient : ^12 ^ 12 -\|1+m\| , -2 < m < 0 . the upper bound on @xmath87 \\| = \\|1 - p_{\mu\nu}\\|$ ] is compatible with the continuum limit , where the @xmath71 s almost commute : @xmath86^\dagger [ t_\mu , t_\nu ] = ( 1-p_{\mu\nu})^\dagger ( 1-p_{\mu\nu})$ ] , @xmath88 , @xmath89 , @xmath90 . any pure gauge action with the right continuum limit and close to it will strongly prefer configurations where all @xmath91 are close to unit matrix . on the other hand , a latticized instanton can be smoothly deformed to a trivial configuration , so there exist gauge field backgrounds for which @xmath92 has an exact zero mode . this is true of any @xmath93 in the range @xmath94 . the above analysis tells us that at least one plaquette will be relatively big for this configuration . unless we enforce the constraint fully , there always will be possibly rare configurations where @xmath26 is not well defined . to fully appreciate the differentiation between the algebraic character of the gw relation and the more complete framework of the overlap , let me give some examples of bad solutions to the gw relation : d^(1)_gw = 12 ( 1+_5 ( h_w ( m_1 ) ) , & d^(2)_gw = 12 ( 1-_5 ( h_w ( m_2 ) ) ; m_1,2 > 0 . + d^(3)_gw = 12 ( 1-_5 ( h_w ( m_3 ) ) ; & -2 < m_3 < 0 .solutions ( 1 ) and ( 2 ) have always trivial index , ( 1 ) has no massless fermions at all , while ( 2 ) has 16 . ( 3 ) is motivated by the @xmath41 symmetry , describes 15 fermions and has a nontrivial index , but 7 fermions contribute to it one way , and 8 the other . the absolute value of @xmath95 is well defined and gauge invariant , but the complex number itself is defined only up to phase . in the general case we have a @xmath96 bundle defined over the space of gauge fields with @xmath97 , a gauge invariant condition . @xmath98 is gauge field independent and gauge invariant and @xmath99 comes from a gauge covariant hamiltonian . thus , the @xmath96 bundle can be viewed as defined over the space of gauge orbits . with the exclusion of the points @xmath100 this space can support non - trivial @xmath96 bundles as required for anomalies . a more familiar way for physicists is to think in terms of berry phases . the main point is that the matrices @xmath101 are smooth in the link variables and this smoothness is inherited by the states @xmath102 . with it come the inevitable berry phase factors . the berry phase factors come from parallel transport on the space of gauge fields with berry s connection @xmath103 . assuming some choice of states @xmath104 ( possibly requiring patches with overlays ) berry s connection is the following one - form over each patch in the space of gauge fields : = v \ { u } \|d v \ { u } . berry s phases come from wilson loop factors associated with @xmath103 . this makes it obvious why they are independent of the choice of phases for the @xmath104 . the non - triviality of berry s phase is locally measured by the associated abelian field strength , a globally defined two - form that usually does not vanish : = da . while @xmath103 depends on the phase choices of the @xmath104 , @xmath105 does not , a fact which is reflected by expressing @xmath105 in terms of the reflection @xmath106 alone : = -14 tr [ ( h_w ( m ) ) d(h_w ( m ) ) d(h_w ( m ) ) ] . to understand the meaning of @xmath103 let us review briefly one way of looking at anomalies in the continuum . we are restricting our attention to the non - abelian case . after all , in the four dimensional abelian case the gauge coupling is nt asymptotically free and the entire issue of mathematical existence of an interacting chiral gauge theory is a moot point . once one has some definition of a regulated chiral determinant one can take a variation with respect to the vector potential @xmath107 to get a current , a nonlocal functional @xmath108 of the gauge field @xmath107 . if the regulated chiral determinant is gauge invariant @xmath108 transforms covariantly under gauge transformations , but when there are anomalies and the regulated chiral determinant is not gauge invariant @xmath108 has a complicated transformation . however , one can always define a covariant current , @xmath109 which does transform covariantly . moreover , @xmath110 is a local , calculable functional of @xmath107 @xcite . one can not use @xmath111 to reconstruct a gauge invariant regulated determinant because it can not be written as the variation of a functional of @xmath107 since it does not obey the wess - zumion consistency conditions . if anomalies do cancel one can find a regularization ( at least in perturbation theory ) with @xmath112 ; then @xmath113 can be regulated first and a regulated gauge invariant determinant can be constructed . the main observation is that berry s connection @xmath103 on the lattice ( when summed over all fermion multiplet contributions ) plays the role of @xmath110 @xcite . to see this we view the lattice currents as one forms over the space of gauge fields . j_consistent = dv^\| v = v^\| d v _ + a . by definition , @xmath114 . the covariant current is j_covariant = v^\| d v _ . and it is quite obvious that it has to transform covariantly because it is unaffected by a phase change in the choice of the @xmath115 and hence is insulated of any gauge breaking step . @xmath103 only depends on the overlap of the state @xmath115 with a state @xmath116 with @xmath117 close to @xmath118 . this is in contrast to the overlap or the currents where @xmath115 is overlapped with a very different state , @xmath98 . this is why @xmath103 is local in the gauge fields but the currents are not @xcite , @xcite . this interpretation of @xmath103 is made solid by the following result : consider an anomalous theory . one can find a two parameter family of gauge backgrounds which becomes a two torus when restricted to orbit space . @xmath105 can also be restricted to orbit space because it is explicitly gauge invariant . one can integrate @xmath105 over this two torus and the result is proportional to an integer . the integer is precisely the same as the one entering the anomaly coefficient . the result can be easily shown to be consistent with a known continuum expression . on the lattice this result means that there are `` monopole '' like singularities in @xmath103 . these singularities can not be removed by small deformations of the matrix @xmath119 . thus , we can not envisage altering @xmath119 so that @xmath103 vanish . but , if anomalies do cancel the obstruction is removed . on the basis of this i propose two conjectures @xcite : * iff anomalies cancel can one deform the total hamiltonian so that one attains @xmath120 without the overlap going through any singularity . * if @xmath120 there is a natural phase choice for @xmath115 , determined by parallel transport with respect to an @xmath103 , such that the action of gauge transformations is non - projective : for any gauge transformation @xmath121 , its representation on the fermions , @xmath72 obeys g(g ) \|v \ { u } = \|v \ { u^g } . if these conjectures are true it would follow that one could fine tune @xmath119 so that , provided anomalies cancel , a natural smooth gauge invariant phase choice for the overlap exists . still , this solution , if it indeed works , looks somewhat contrived for nature because it requires a fine tuning of the hamiltonian used in the construction . i believe that this fine tuning is not necessary . if a fine tuned hamiltonian exists , any hamiltonian close enough to it would also work , although a residual gauge dependence would be present . one would simply average over the gauge degrees of freedom ( integrate them out ) and this would have no effect on the continuum limit @xcite . if anomalies do not cancel , the gauge dependence is always large enough that gauge averaging induces new non - local terms in the action . indeed , the gauge dependence contains a lattice version of a nontrivial wess - zumino term . when anomalies cancel but one still has a small amount of gauge breaking , gauge averaging only adds some extra gauge invariant local terms to the action . this has been shown numerically in a two dimensional chiral model @xcite . there exists an important special case where there are anomalies but @xmath105 vanishes . it is in four dimensions with gauge group @xmath122 and weyl fermions in the @xmath123 representation . then the matrix @xmath124 can be made real by a global base choice . still , berry s phase factors can take the values @xmath125 and thus witten s anomaly is reproduced @xcite . this again shows that in the overlap all anomalies are encoded in berry s phases . although the overlap at present does not provide a rigorous construction for chiral gauge theories in the non - abelian case , it has passed the boundary of physical plausibility , relegating the completion of this construction to the branch of mathematical physics . for physics the main question is : does there exist a truly new way to regulate chiral symmetries , or is the way opened by kaplan and by frolov and slavnov and subsequently realized by the overlap unique in some sense ? my research at rutgers is partially supported by the doe under grant # de - fg05 - 96er40559 . i wish to express my appreciation of the immense hospitality and great effort invested by the organizers of chiral 99 in taipei . in particular i wish to thank ting - wai chiu for doing so much to produce an inspiring and enjoyable meeting . 99 h. neuberger , hep - lat/9909042 . s. adler , ; j. s. bell , r. jackiw , ; s. adler , w. bardeen , . b. zumino , `` chiral anomalies and differential geometry '' les houches 1983 , proceedings , relativity , groups and topology , ( qc6:g7:1983 ) ; r. stora , `` continuum gauge theories '' , cargese summer institute 1976 , ( qc174.45:c35:1976 ) . r. d. ball , ; l. alvarez - gaume , `` an introduction to anomalies '' , erice school math.phys.1985:0093 ( qcd161:i569:1985 ) . c. callan , j. harvey , . d. boyanowski , e. dagotto , e. fradkin , . d. b. kaplan , . s. a. frolov , a. a. slavnov , . r. narayanan , h. neuberger , . r. narayanan , h. neuberger , . r. narayanan , h. neuberger , . r. narayanan , h. neuberger , ; ; . y. kikukawa , r. narayanan , h. neuberger , ; . s. randjbar - daemi , j. strathdee , , ; ; ; . h. neuberger , . p. ginsparg , k. wilson , . m. l " uscher , hep - lat/9909150 . h. neuberger , . h. neuberger , . r. narayanan , . t. w. chiu , s. v. zenkin . t. kato , `` perturbation theory for linear operators '' , springer - verlag , berlin , 1984 . j. avron , r. seiler , b. simon , . s. borac , r. narayanan , h. neuberger , . t. w. chiu , hep - lat/9910029 . h. neuberger , hep - lat/9911004 . w. bardeen , b. zumino , h. neuberger , . s. randjbar - daemi , j. strathdee , . d. f " orster , h. b. nielsen , m. ninomiya , . h. neuberger , ; .	the overlap formulation of regulated vectorial and chiral gauge theories is reviewed . ostensibly new constructions , based on the ginsparg - wilson relation are essentially just overlap with new notation . at present there exists no satisfactory realization of chiral symmetries outside perturbation theory which is structurally different from the overlap . # 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) 11.15.ha , 12.38.gc
one of the most intriguing questions in modern astronomy is how the universe assembled into the structures we see today . a major part of this question is understanding how galaxies formed over cosmic time . a popular and rapidly developing method of tracing the evolution of galaxies is through examining galaxy morphologies and structures over a range of redshifts ( e.g. conselice 2003 ; conselice , rajgor & myers 2008 ; conselice et al . 2009 ; lotz 2008 ; cassata et al . 2010 ; ricciardelli et al . 2010 ; weinzirl et al . while understanding how the morphologies of galaxies evolve , and how matter in galaxies is structured , is fundamental , structural studies of galaxies have thus far focused on measuring properties in one or more photometric bands , and using this as a tracer of evolution . these types of structural analyses have always been measured in terms of relative luminosities , such that brighter parts of galaxies contribute more towards their structure and morphology . to understand galaxies more fully however , it is important to study the evolution of galaxy structure in terms of their stellar mass distribution , as this better reflects the underlying distribution of stars in a galaxy . morphological studies have evolved from initial attempts to describe the range of galaxy forms during the early - mid 20th century , towards modern efforts of linking the spatial distribution of a galaxy s stars to its formation history . in the move away from visual classifications , the goal has been to quantify galaxy structure / morphology in a way such that structures can be measured automatically and in a reliable way . there are two broad approaches for automated classification of galaxies - the parametric and non - parametric methods . in the parametric approach , the light profiles of galaxies , as seen in the plane of the sky , are compared with predetermined analytic functions . for example , single sersic profiles are fit to an entire galaxy s light profile . similarly , bulge - to - disc ( @xmath8 ) light ratios are computed by fitting two component profiles to a galaxy . these methods are however unable to parameterise directly any star formation or merging activity which produces random or asymmetric structures within galaxies . the `` non - parametric '' structural approaches have no such presumed light profiles implicit in their analyses ( e.g. , shade et al . 1995 ; abraham et al . 1996 ; conselice 1997 , 2003 ; lotz et al . this is a more natural approach towards measuring galaxy structures , as the majority of galaxies in the distant universe have irregular structures that are not well fit by parameterised forms ( e.g. conselice 2005 ; conselice , rajgor & myers 2008 ) . perhaps the most successful and straightforward of the non - parametric systems is the @xmath9 method , which uses a combination of concentration ( @xmath10 ) , asymmetry ( @xmath11 ) and clumpiness ( @xmath12 ) values to separate galaxy types ( conselice 2000 ; bershady 2000 ; conselice 2003 ) . besides overall structure , the measurement of the size evolution of galaxies , as measured through half - light radii , also has important implications for their formation histories . for example , many studies such as trujillo ( 2007 ) and buitrago et al . ( 2008 ) have measured galaxy sizes for systems at @xmath13 and have found a population of compact spheroid galaxies with number densities two orders of magnitude higher than what we find in the local universe . the absence of a significant number of these small sized ( as measured by half - light radii ) , high mass galaxies in the local universe suggests that this population has evolved and has perhaps merged with other galaxies . however , these studies are measured based on the light originating from these galaxies , and it is not clear whether sizes would change when measured using the distribution of stellar mass rather than light . all of these methods for measuring the resolved structures of high redshift galaxies depend on measurements made in one or more photometric bands . as such , quantitative structures are influenced by a combination of effects , including : regions of enhanced star formation , irregular dust distributions , differing ages and metallicities of stellar populations and minor / major mergers . the morphology and structure of a galaxy also depends strongly upon the rest - frame wavelength probed ( e.g. , windhorst et al . 2002 ; taylor - mager et al . young stars , such as ob stars can dominate the appearances of galaxies at blue wavelengths , thereby giving a biased view of the underlying mass within the galaxy . longer wavelength observations improve the situation , but at every wavelength , the light emitted is from a mixture of stars at a variety of ages and the dust content . more fundamental is the structure of the stellar mass distribution with a galaxy , as it is more of a direct tracer of the underlying potential . although galaxy structure has been measured on rest - frame near - infrared and @xmath14-band imaging , which traces stellar mass to first order , we directly investigate the stellar mass images within this paper . we use the method outlined in lanyon - foster , conselice & merrifield ( 2007 ; lcm07 ) to reconstruct stellar mass maps of galaxies within the goods fields at @xmath0 . we then use this to investigate the evolution of the distribution of galaxy stellar mass during the last half of the universe s history . we use these stellar mass maps to directly measure @xmath9 parameters , and the sizes of these galaxies in stellar mass over this epoch . we test the assumptions that the @xmath9 parameters and sizes can be reliably measured in optical light by comparing the parameters for these galaxies measured in @xmath15 , @xmath16 , @xmath17 , and @xmath6 bands and in stellar mass . we finally investigate how these various wavelengths and stellar mass maps can be used to classify galaxies by their formation modes , revealing how these systems are assembling . this paper is organised as follows : in [ sec : data ] and [ sec : method ] we describe the data , sample selection and method , including explanations of the k - correction code we use and the @xmath9 analysis . in [ sec : results ] we present our results , discussing the stellar mass maps themselves , the comparison between galaxy size and the @xmath9 parameters in @xmath6 and mass . we then explore the relations between the structural parameters in @xmath18 , @xmath16 , @xmath17 , @xmath6 and stellar mass . finally , in [ sec : conclusions ] , we discuss our conclusions and comment on future applications . throughout we assume a standard cosmology of _ km s@xmath20 mpc@xmath20 , and @xmath21 = 0.3 . the primary source of our data consists of hst / acs imaging from the goods acs imaging treasury program . the observations consist of imaging in the @xmath18 ( f435w ) , @xmath16 ( f606w ) , @xmath17 ( f775w ) and @xmath6 ( f850lp ) pass - bands , covering the hubble deep field - north ( hdf - n ) area . the central wavelengths of these filters , and their full - width at half - maximum , are : f435w ( 4297 , 1038 ) , f606w ( 5907 , 2342 ) , f775w ( 7764 , 1528 ) , f850l ( 9445 , 1229 ) . the images have been reduced ( using the acs calacs pipeline ) , calibrated , stacked and mosaiced by the goods team ( giavalisco 2004 ) . the field is provided in many individual image sections , with the hdf - n data divided into 17 sections , each of @xmath22 pixels . the total area of the goods survey is roughly 315 arcmin@xmath23 , and we utilise imaging which was drizzled with a pixel scale of 0.03 arcsec pixel@xmath20 , giving an effective psf size of @xmath24 in the @xmath25band . the sample consists of galaxies selected in the goods - n field with @xmath26 and a f814w magnitude of @xmath27 . the sample was also restricted to those galaxies with data in all four acs photometric bands such that we can subsequently obtain the most accurate k - corrections and stellar masses possible . the final sample consists of 560 objects , which were individually cut out of the goods - n acs imaging , and handled as separate entities . this is described in detail in the following section . photometric redshifts are available for the whole sample ( mobasher 2004 ) , and spectroscopic redshifts are available for 404 out of 560 galaxies , as found by the team keck redshift survey ( tkrs ; wirth , 2004 ) . when spectroscopic redshifts are available they are used , else the photometric values are substituted . to create stellar mass maps of our galaxies , each pixel in each galaxy image is treated individually throughout the analysis . we convert fluxes from counts per pixel to apparent magnitudes per square arc - second for each pixel in every image in all four photometric bands , as well as calculating their associated errors . pixels with negative fluxes were assigned values four orders of magnitude smaller than the typical flux so that even pixels with low signal - to - noise values can be mapped . this however means that our resulting values for the cas parameters , especially asymmetry which uses the background light to do a correction , are potentially different than when measured using optical light . because our galaxies are at a variety of redshifts , we have to carry out fitting to each sed for each pixel to calculate the stellar mass within each pixel . k - corrections and stellar masses are calculated for each pixel of each image using the `` k - correct '' code of blanton & roweis ( 2007 ) . this is similar to previous methods outlined by lanyon - foster et al . ( 2007 ) , bothun ( 1986 ) , abraham et al . ( 1999 ) , and welikala et al . ( 2008 , 2011 ) . the code , k - correct naturally handles large datasets , and through its interpretation of the data in terms of physical stellar population models , outputs results in terms of stellar mass and star formation histories ( sfhs ) of each galaxy . the code allows the input of goods data and outputs a stellar mass for each object . the code can also be modified so as to treat each pixel as a separate input object , which we do in this work . the k - correction ( @xmath28 ) between bandpass @xmath29 , used to observe a galaxy with apparent magnitude ( @xmath30 ) , at redshift @xmath31 , and the desired bandpass @xmath32 is defined as ( e.g. , oke & sandage 1968 ) : @xmath33 where @xmath34\ ] ] is the bolometric distance modulus calculated from the luminosity distance , @xmath35 , @xmath36 is the absolute magnitude and @xmath37 = h@xmath38/100 kms@xmath20mpc@xmath20 . the k - correct software contains model templates in electronic form and an implementation of the method to fit data to models . the code uses the stellar population synthesis models of bruzual & charlot ( 2003 ) and contains training sets of data from galex , sdss , 2mass , deep2 and goods . the code finds the nonnegative linear combination of @xmath39 template star formation histories that best match the observations using a minimum @xmath40 comparison . with the entire set of galaxy observations available , k - correct also fits for the @xmath39 template sfhs using a nonnegative matrix factorisation algorithm . k - correct naturally handles data uncertainties , missing data , and deals with the complications of observing galaxy spectra photometrically , using broadband filters of galaxies at varying redshifts . we use a set of 485 spectral templates to fit to our galaxy pixel seds using bruzual & charlot ( 2003 ) models with the chabrier ( 2003 ) imf and padova ( 1994 ) isochrones . all of the six metallicities available ( mass fractions of elements heavier than he of @xmath41 and @xmath42 ) are used . some of the known areas of uncertainty in the stellar population models are in the uv and ir regions . in the uv light from young or intermediate aged stellar populations can dominate the flux at @xmath43 . in the near - ir , thermally pulsating asymptotic giant branch ( tp - agb ) stars dominate the flux in some intermediate age populations ( maraston , 2005 ) . we discuss this issue in terms of stellar masses in conselice et al . ( 2007 ) who show that the effects of tp - agb stars do not kick in until at redshifts higher than the scope of this study ( @xmath44 ) , and thus are not a concern in the present work . results are outputted from k - correct for each input pixel , giving rest - frame fluxes at @xmath45 , @xmath46 , @xmath47 , @xmath48 , @xmath29 , @xmath14 magnitudes plus their associated errors , mass - to - light ratios in all of these bands and finally the stellar mass in that pixel . we then reconstruct the image of the galaxy in stellar mass based on these calculations across the galaxy . we use a signal - to - noise limit of s / n @xmath49 per pixel in stellar mass for all optical bands , for a reliable calculation . we also investigate how the results would change if pixels are summed together , finding similar results with the exception of the asymmetry which tends to decrease at lower resolutions . images and segmentation maps.,width=321,height=321 ] we have classified our entire sample visually , using the @xmath6 galaxy images and the object segmentation maps . we have derived the classification system by applying the most appropriate criteria to this particular dataset . the classification scheme is divided into nine categories , ranging from compact objects ( spherical with little or no apparent envelopes ) to obviously merging systems ( with evidence of tidal streams and other merger attributes ) . six objects in the catalogue were found to be unresolved and these were removed from the sample . we give a description of these types and how they were selected below . we often refer to these galaxy types throughout this paper , and appendix b gives a description of these types in terms of our measured indices . figures 2 - 10 show examples of these galaxy types . , left ) , stellar mass ( middle ) and the distribution of @xmath50 ratios ( right ) for the compact elliptical ( ce ) galaxies . the @xmath50 map is scaled such that white corresponds to higher @xmath50 values , and black to lower @xmath50 values . this convention is used in the following figures [ fig : massmap_pe ] to [ fig : massmap_m].,width=321,height=491 ] ) , stellar mass maps , and the distribution of @xmath50 ratios for examples of peculiar elliptical ( pe ) galaxies . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] the _ compact ellipticals _ , denoted as `` ce '' , were identified as compact objects that varied smoothly across their radii when viewed morphologically . these objects had little or no sign of an envelope usually associated with early - type galaxies . the compact es are generally found at lower redshifts , at @xmath51 with a few between @xmath52 . the distinguishing feature of these compact es is their small sizes , with half - light radii of @xmath53 kpc . the elliptical galaxies are amongst the most popular and important galaxy type in our study , as they are typically the most massive galaxies in the local universe . one reason for this is that elliptical galaxies , and massive galaxies in general , are the test - bed for understanding theories of galaxy formation which often have strong predictions for how the most massive galaxies should form ( e.g. , bertone & conselice 2009 ) . the ellipticals we see at high redshift are likely the progenitors of the massive ellipticals found today . galaxies with the familiar features of nearby _ elliptical galaxies _ were classed as such , and given the label e " . 28 e galaxies were identified , which at first glance have the appearance of early types , but on closer inspection and by varying the image contrast had some peculiar features , such as multiple nuclei or minor disturbances in the internal structure . these objects were classified as _ peculiar ellipticals _ , and are labelled as pe ( conselice 2007 ) . objects with a disc - plus - bulge structure were classified as _ early - type spirals _ ( es " ) if the bulge appears larger than the disc component , and _ late - type spirals _ ( ls " ) if the disc is larger than the central bulge . edge - on disc galaxies are denoted by eo " . the spiral galaxies are amongst the most interesting for this study given that they often contain two major stellar population types segregated spatially . traditionally this is seen as an older stellar population making up the bulge , and the spiral arms consisting of younger stellar populations . objects that did not fit easily into any of the previously defined categories , but whose segmentation map showed them to be singular , or unconnected to any apparently nearby object in the image , were classified as _ peculiar _ , pec " . these are galaxies that could possibly have merged in the recent past . galaxies whose morphology also ruled them out from any previously defined class but whose segmentation map showed them to be connected to , or associated with , at least one other object on the image , but without obvious merger signatures , were classified as _ possible mergers _ ( pm " ) . galaxies with the pm requirements and which showed obvious signs of merging , such as tidal tails , were classified as _ mergers _ ( m " ) . this information is summarised for quick reference in table [ tab_classfcn ] . the sample was independently classified four times . the mode of the results were taken to be the true classification and where an indecisive split was produced the lowest numerically valued type was chosen . figure [ fig : class ] shows a histogram of the sample classifications . we note that there are a significant number of peculiar galaxies in our sample ( e.g. , figure 1 ) , much higher than what has been found at lower redshifts in previous work investigating galaxy morphology ( e.g. , conselice et al . 2005 ) . these are systems that at lower resolution would be classified as disk - like galaxies in many cases . they are not necessarily merging systems , but are likely normal galaxies in some type of formation , the modes of which are one of the features we investigate later in this paper . we show examples of each of the above types in figures 2 - 10 , with five examples shown for each . these figures show the image of the galaxy in the @xmath6-band , the stellar mass maps , and the mass to light ratio for each galaxy . .descriptions of the classification scheme . [ cols="<,<",options="header " , ] for the elliptical ( e ) galaxies . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] for the early - type spiral galaxies ( es ) . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] for the late - type spiral galaxies ( ls ) . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] for the edge - on disk galaxy ( eo ) galaxies . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] for the peculiar ( pec ) galaxies . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] for the possible - merging ( pm ) galaxies . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] for the merging ( m ) galaxies . the scale and convention is the same as that used in fig . [ fig : massmap_ce].,width=321,height=491 ] we use the concentration , asymmetry , clumpiness ( @xmath9 ) parameters to quantitatively measure the structures of our sample , in all available bands , @xmath2 , and on the stellar mass maps . we also measure the gini and m@xmath54 parameters , forming an extensive non - parametric method for measuring the structures and morphologies of galaxies in resolved ccd images ( e.g. , conselice et al . 2000a ; bershady et al . 2000 ; conselice et al . 2002 ; lotz et al . 2004 ; conselice 2003 ; lotz et al . the premise for using these parameters is to tap into the light distributions of galaxies , which reveal their past and present formation modes ( conselice 2003 ) . the regions into which the traditional hubble types fall in cas parameter space is well understand from local galaxy comparisons . for example , selecting objects with @xmath55 finds systems that are highly disturbed , and nearly all are major galaxy mergers ( e.g. , conselice et al . 2000b ; conselice 2003 ; hernandez - toledo et al . 2005 ; conselice 2006a ) . a more detailed analysis of this problem is provided in appendix a for optical light . ) colour vs. the magnitude of the galaxy as observed in the @xmath6 band . different galaxy types are displayed , including : compact ellipticals ( ce ) , peculiar ellipticals ( pe ) , ellipticals ( e ) , early - type spirals ( es ) , late - type spirals ( ls ) , edge - on disk galaxies ( eo ) , peculiar galaxies ( pec ) , possible - merger galaxies ( pm ) and merging systems ( m ) . this key is used throughout the remainder of this paper.,width=340,height=340 ] we measure the structural parameters for the goods sample using the method of conselice ( 2008 ) , with slight adjustments made for the stellar mass maps , to enable the code to handle the large values of stellar masses per pixel , rather than flux . the radius of each individual galaxy within the postage stamp image / mass map is measured on the stellar mass map , and we define all our indices within the petrosian radii ( e.g. , petrosian 1976 ; bershady 2000 ; conselice 2003 ) . the petrosian radius has been found to be a better are more reliable radius ( and reproducible ) than the isophotal radius ( petrosian 1976 ) . the limits and relationship to other radii is described in detail in terms of total light in a galaxy by graham et al . ( 2005 ) . circular apertures are used for measuring our petrosian radii and quantitative parameter estimation . the petrosian radius used to measure our parameters is defined by , @xmath56 where @xmath57 is the radius where the surface brightness ( or stellar mass per unit area ) is 20 percent of the surface brightness ( or stellar mass per unit area ) within that radius ( bershady et al . note that this is a distance independent measurement , given that surface brightness dimming effects both measurements of surface brightness in the same way . accounting for background light and noise is extremely important when measuring structural parameters , especially for faint galaxies , and this must also be dealt with for the stellar mass maps . the measured parameters in the @xmath18 , @xmath16 , @xmath17 and @xmath6 bands and stellar mass images are corrected as described in conselice ( 2008 ) , by considering a background area close to the object , and the segmentation map of the object . using a background close to the galaxy itself , any problems introduced by objects imaged on a large mosaic , with a non - uniform weight map , and the object itself being faint compared to the background , are alleviated . we review below how the @xmath9 parameters are measured . these are described in more detail in bershady et al . ( 2000 ) , conselice et al . ( 2000 ) , conselice ( 2003 ) and lotz et al . ( 2008 ) . we measure the asymmetry of a galaxy by taking the original galaxy image ( or stellar mass map ) , rotating it by 180 degrees about its centre , and then subtracting the two images ( conselice 1997 ) , with corrections for background and radius ( see conselice 2000a for details ) . the centre of rotation is found using an iterative process , which locates the minimum asymmetry . a correction is made within the stellar mass maps for this parameter , as it was found that uncertainties in the mass - to - light ratio adversely affected the asymmetry measurement . we found that approximately 15 percent of the stellar mass in the asymmetry calculation is left over from random fluctuations in @xmath50 . this was compensated for including an addition correction for this uncertainty through a slight additional subtraction to the asymmetry signal . this was done by scaling the asymmetric residuals of the rotated and subtracted image by an increase of 15% , that is only 85% of the residual is considered . we measure the background light in the same way as for the light measures of asymmetry , although because we are dealing with a conversion to stellar mass , which is not trivially done for the background , we have some background values which are lower than zero . to deal with this , we placed all the negative pixels in the stellar mass map background to zero . the equation for calculating asymmetry is : @xmath58 where @xmath59 denotes the original image pixels , @xmath60 is the image after rotating by 180 . the background subtraction is made using light ( or mass ) from a blank sky area , @xmath61 , and is minimised using the same process as for the object itself . the higher the values of @xmath11 , the higher the degree of asymmetry the galaxy possesses , the most extreme cases usually corresponding to merger candidates ( conselice 2003 ) . the concentration parameter measures the the intensity of light ( or stellar mass ) contained within a pre - defined central region , compared to a larger region towards the edge of the visible galaxy . concentration is most often defined as the ratio of the flux contained within circular radii possessing 20 percent and 80 percent ( @xmath62 , @xmath63 ) of the total galaxy flux , @xmath64 a higher value of @xmath10 corresponds to an object where more light is contained within the central region . this measurement has been shown to correlate with the halo , and total stellar masses , of nearby galaxies ( e.g. bershady 2000 ; conselice 2003 ) . ) vs. redshift ( @xmath31 ) for our sample . the points displayed are the same as in fig . [ fig : cmd].,width=340,height=340 ] clumpiness ( @xmath12 ) is related to asymmetry , in that it is used to measure the amount of light ( or mass ) in a galaxy that exists in discrete , clumpy distributions . in this definition , a smooth galaxy contains light at low spatial frequencies , such as elliptical galaxies , whereas clumpy systems have most of their light contained in high spatial frequencies . star forming galaxies tend to be clumpy in structure , with high @xmath12 values . we measure clumpiness via : @xmath65,\ ] ] where , the original image @xmath66 is blurred to produce a secondary image , @xmath67 . the secondary image is subtracted from the original image to create a residual map , showing only the high frequency structures contained within the galaxy ( conselice 2003 ) . the residuals are quantified by normalising , using the total light in the original galaxy image , and then subtracting the normalised residual sky . the smoothing kernel , @xmath68 , used is determined from the radius of the galaxy , and has the value @xmath69 ( conselice 2003 ) . the centres of the galaxies ( roughly the inner 1/10th the petrosian radius ) are removed during this procedure . this parameter is furthermore the most difficult to measure and is best measured on galaxies imaged with a high s / n ratio , and are only of limited use in this paper . also , the clumpiness is highly sensitive to resolution or seeing , although the acs imaging we use at these redshifts is within the range where values can be properly compared to nearby calibration data sets ( conselice 2003 ) . the gini coefficient ( @xmath70 ) is defined through the lorentz curve of the light distribution and does not depend on a predefined central position , thus distinguishing it from the concentration parameter . if @xmath70 is zero , the galaxy has a uniform surface brightness , whereas if a single pixel contains all the flux , @xmath70 is unity . to calculate @xmath70 efficiently , the pixel flux ( or stellar mass ) values , @xmath71 , are sorted into increasing order and the following formula is used : @xmath72 where @xmath73 is the total number of pixels in the galaxy ( lotz 2004 , 2008 ) . m@xmath54 is anti - correlated with concentration such that galaxies with high concentration have low m@xmath54 values . m@xmath54 is a direct tracer of the brightest regions in a galaxy , but does not require a pre - determined central coordinate , rather this is calculated to be the location of minimal m@xmath54 . m@xmath54 is more sensitive to merger signatures than concentration and is normalised by the total moment of the galaxy . it is defined as @xmath74 @xmath75 \label{eq : mtot}\ ] ] where @xmath76 , @xmath77 is the galaxy centre ( lotz 2004 ) . one issue that we must address within this paper is how well we are measuring structure on the stellar mass maps for our sample of galaxies , and importantly how the measurement of stellar mass structure would change with redshift due only to cosmological effects . this is critical if we are to make any inferences of the evolution of galaxies in terms of their stellar mass distribution over time . this issue has been addressed before for visual images of galaxies in conselice ( 2003 ) and in beauvais & bothun ( 1999 ) for velocity fields in spiral galaxies . here we address it for the stellar mass maps . we simulate how our stellar mass map creation process , and cas measurements would vary as a function of redshift due to a decrease in the signal to noise and resolution . we carry out this process in the same way we have for the actual galaxies in the goods fields that we analyse in this paper . we take a sample of 82 nearby galaxies of various types - ellipticals , spirals , and irregulars , convert their flux values to stellar mass . we also go through the entire process we do on goods , including setting equal to zero the background values with negative stellar masses . this stellar mass conversion is done after the galaxy is redshifted , so that we are mimicking as much as possible how the real observations and measurements are done . the simulation itself is done by simulating each image at some redshift z@xmath78 to how it would appear at z@xmath79 , with z@xmath80 z@xmath78 . in our case z@xmath81 . when carrying out these simulations of placing lower redshift galaxies to high redshifts we calculate first the rebinning factor , b , which is the reduction in apparent size of a galaxy s image when viewed at higher redshift . the other major factor is the relative amounts of flux from the sky and galaxy , and the noise produced from the galaxy , sky , dark current , and imaging instrument ( e.g. , read noise ) from acs . the process and details for how these simulations are done can be found in conselice ( 2003 ) . in summary , the surface brightness of the simulated image must be reduced such that the equation , @xmath82 holds , whereby the galaxy as observed in one filter , and is simulated in another with central rest - frame wavelengths of @xmath83 and @xmath84 and widths @xmath85 and @xmath85 . in the above equation @xmath86 is the total number of pixels within the galaxy at @xmath31 and p@xmath87 is the average adu counts per pixel . the calibration constant @xmath88 is in units of erg s@xmath20 @xmath89 @xmath20 adu@xmath20 . a sky background , and noise from this backgrounds is added to these images by @xmath90 , where @xmath91 is background flux in units of in units of adu s@xmath20 . for acs simulations we take these values from the measured background based on goods imaging ( giavalisco et al . 2004 ) checked to be consistent with the values from the acs handbook . other noise effects are then added , including read - noise scaled for the number of read - outs , dark current and photon noise from the background . our resulting images are then smoothed by the acs psf as generated by tiny - tim ( kriss et al . 2001 ) , although using psfs measured from stars give the exact same results . from these simulated images at redshifts z = 0.5,1 and 2 , we then reconstruct the stellar mass images in the same way as we do for our original galaxies , and then measure the cas parameters on these stellar mass images . the results of this are shown in figure [ fig : sim ] for the concentration and asymmetry parameters . the m@xmath54 and gini parameters however have similar behaviours , as do the galaxy half - light radii . we find that the clumpiness indices have a similar average difference , but with a much larger scatter . what we find overall is that the trend with redshift is such that the concentration and asymmetry parameters on average do not change significantly when measured in stellar mass ( figure [ fig : sim ] ) . we show in the right panel of figure [ fig : sim ] the change in c and a when divided into early / late / peculiar types . again , on average there is not a significant change with redshift , although individual galaxies clearly can have significant differences at higher redshift . we also investigate how our parameters change at each redshift between the stellar mass and the visual images after the simulation . we find very little difference except that at the highest redshifts of our simulations , at @xmath92 , we find that the average change in the asymmetry parameter changes @xmath93 , such that the stellar mass image is more asymmetric than the optical light one . we also investigate the same quantities that we use in later figures , finding that twice the difference in the asymmetries of the optical and stellar mass image , divided by the sum of these , increases steadily until it reaches a values of @xmath94 at @xmath92 similar to what we see later in this paper for the actual values . however at @xmath95 we find that this change is less and closer to @xmath96 . the following sections describes the analysis of our sample , and what we can learn from examining their structures in stellar mass maps , and how this evolves over time . we first examine the distribution of visual classifications for our sample and their global colours . we later describe the stellar mass maps we construct for these galaxies , and then finally present a structural analysis of these systems based on the distribution of their stellar mass . in figure [ fig : cmd ] we show the observed colour magnitude diagram for our sample . each morphological class is plotted as a different symbol ; solid red circles represent the ce s , open red circles show the pe galaxies and the e galaxies are represented by red stars . solid blue squares show the es and open blue squares the ls systems . edge - on disk galaxies are represented by black diagonal crosses and solid green triangles show the pec systems . pm and m galaxies are represented by open green triangles and open green stars respectively . this key is used throughout the rest of the paper in the various figures . based on this , we find that our classified early - type galaxies are on average redder and brighter than the peculiars and pm / m systems , while the spirals tend to be bluer across a range of brightnesses , but the differences are not so clear as would be expected for a sample at low-@xmath31 only . this colour - magnitude diagram does not show such an obvious morphological sequence as is seen for purely hubble type galaxies in the nearby universe ( e.g. , conselice 2006b ) . the relation between @xmath97 colour and redshift is shown in figure [ fig : colz ] . the symbols are the same as those in fig . [ fig : cmd ] . as can be seen , there is a general trend for all types to become redder with increasing redshift , which is an expected feature of redshift . the early types are reddest across the whole redshift range , followed by the es and ls galaxies , with the pec / pm / m systems the bluest across all redshifts . the compact ellipticals are blue and faint . note however that some pm and es galaxies are quite red . what this reveals is that there is a decoupled relation between the colour and morphologies of galaxies not seen in nearby galaxies where this correlation is strong . a specific morphological type as measured visually at high - z can not be used to predict the colour of the galaxy . by calculating stellar mass to light ratios and the stellar mass within each pixel of each galaxy image in our sample , we reconstruct the image of each galaxy in terms of stellar mass and mass - to - light ratio . we do this for each galaxy , using the method described in lcm07 , and present five representative examples for each classification in figures [ fig : massmap_ce ] to [ fig : massmap_m ] . this figures present galaxy images in @xmath6 ( left ) , stellar mass ( centre ) and @xmath98 ( right ) , which we discuss below . the images are grey - scaled such that the darkest pixels represent those brightest in @xmath6 , and those that are most massive . in the mass - to - light image , the whitest pixels are those with the highest ( @xmath50 ) , or redder in colour . we explain below how these maps appear for our various types . figure [ fig : massmap_ce ] shows five examples of the compact elliptical ( ce ) population in the sample . these galaxies appear similar in their stellar mass maps as they do in their @xmath6 images at first glance . the mass - to - light maps are not as uniform , and there are important quantitative differences between the stellar mass and light images . for example , the galaxy 18246 appears to have a relatively higher ( @xmath50 ) in its core compared to its outer parts , whilst galaxy 25035 appears to have a low ( @xmath50 ) in its centre , suggesting that it has a bluer core , likely due to star formation . galaxy 36503 contains a high ( @xmath50 ) ring surrounding a blue core . five examples of the peculiar elliptical ( pe ) galaxies are displayed in fig . [ fig : massmap_pe ] . these galaxies appear more diffuse in stellar mass than in @xmath6 , in part due to the loss in contrast in the stellar mass maps , but also due to their blue inner colours ( fig . [ fig : massmap_pe ] ) . as can be seen in the ( @xmath50 ) ratio maps of these galaxies , they have blue cores , and the effect is most pronounced in 22932 and 27429 . these galaxies also have blue overall colours compared to pure e galaxies ( fig . [ fig : colz ] ) . these peculiar ellipticals have been seen and studied before in papers such as conselice et al . they generally have a high asymmetry , and are found amongst the most massive galaxies in the universe at @xmath99 . as can be seen in the m / l maps for these systems ( fig . [ fig : massmap_pe ] ) , these galaxies also have a diversity in how young and old stellar populations , including dust , are distributed in these systems compared to normal elliptical galaxies . figure [ fig : massmap_e ] presents five examples of early - types from our sample . these galaxies appear more diffuse and more distributed spatially in their stellar mass than in the @xmath6 band , especially at large radii . these galaxies have larger ( @xmath50 ) ratios in their centres , except for 36419 , which is blue , and has a more complicated structure in ( @xmath50 ) than the others . overall , the e galaxies morphologically have very similar structures in stellar mass and @xmath6 . figures [ fig : massmap_es ] and [ fig : massmap_ls ] show the @xmath6 , stellar mass and ( @xmath50 ) maps for the early - type , and late - type spirals . as found in lcm07 , structures within discs , including prominent spiral arms , are often ( but not always ) smoothed out in stellar mass . this is true for early types as well as late - type spirals , as can be seen for example in galaxy 25465 ( fig . [ fig : massmap_es ] ) . inspection of the ( @xmath50 ) maps reveal that the discs of many spirals are bluer than their bulges , as expected . the mass - to - light ratio maps of the edge on galaxies ( fig . [ fig : massmap_eo ] ) are mostly homogeneous in structure . however , the maps of 19280 , 39312 and 49722 all show some patches of low @xmath50 . indeed , 19280 appears to be blue across the whole image , which would imply that the dust content of this galaxy is low , which is unusual for edge - on galaxies . figure [ fig : massmap_pec ] shows the f814w images , stellar mass maps and m / l ratio maps for the peculiar galaxies within our sample . the peculiar galaxies 22690 and 25584 ( fig . [ fig : massmap_pec ] ) appear to contain objects near the primary galaxy , but in both cases it is the larger galaxies on the right that is the target . although 22690 looks disturbed in @xmath6 , the effect of smoothing in the stellar mass map is also seen here . this indicates that disturbed regions of the galaxy are due to star forming regions . the @xmath6 image of 22690 does not have the regular morphology of a nearby spiral galaxy , although the stellar mass map has a structure one would expect of such a galaxy , having a central bulge surrounded by a smooth disc . the ( @xmath50 ) map also shows a red central region surrounded by bluer pixels . the mass - to - light maps of the peculiars generally show that they are blue and also difficult to see in stellar mass , due to their low mass to light ratio . this suggests that star formation itself might be difficult to trace in stellar mass . however , the stellar mass in the pm galaxy 18917 ( fig . [ fig : massmap_pm ] ) traces the light in the galaxy , despite having a low ( @xmath50 ) . note also that the peculiars are often peculiar spirals , in the sense that they look like nearly normal spirals , but with some peculiar features . often these peculiars are quite small as well , and many of them are very likely spirals in some type of formation . the merging galaxy system 21839 ( fig . [ fig : massmap_m ] ) is similar in stellar mass to its @xmath6 image , but possesses a more intricate structure in ( @xmath50 ) . the mass - to - light map shows the smaller merging object to be blue , whereas the brighter galaxy appears more red but with blue regions in its core . the system 29800 ( fig . [ fig : massmap_m ] ) , whilst clearly appearing to be a merger between two galaxies of approximately equal brightness in @xmath6 , would likely be classified as a disc galaxy in stellar mass . the bottom of the two bright nuclei has disappeared completely in the stellar mass image , although it is visible as a low ( @xmath50 ) patch in the mass - to - light map . in this section we compare how the distribution of stellar mass in a galaxy compares to the distribution of light as seen in the @xmath6 acs imaging . one of our goals is to determine how appropriate studies in @xmath6 and similar red bands are for measuring the mass content and structure of a galaxy , and how stellar mass quantitative morphologies differ from those measured in light . in this section we investigate the relationship between galaxy properties in stellar mass and @xmath6-band light . for galaxy size ( as measured by the half - light radius ) and the @xmath9 parameters we plot the normalised difference between @xmath6 and stellar mass versus the mean value of @xmath6 and stellar mass as a representative figure to determine how these various quantities change . one of the basic features of a galaxy is its size , which in this paper we quantitatively mean the half - light radii . galaxy half - light radii are found to strongly evolve with time , such that galaxies at higher redshifts have a more compact structure ( e.g. , ferguson et al . 2004 ; trujillo et al . 2007 ; buitrago et al . 2008 ; carrasco et al . however , every size measurement has been carried out through measurements of light , and it is desirable to determine how the half - mass radii of galaxies measured in stellar mass maps compares with half - light radii measured using light . as such , we have calculated half - light ( or half - mass ) radii ( @xmath100 ) , in kpc , for our sample in both stellar mass and @xmath6 light , and investigate whether half - light / mass radii are comparable . we define the normalised size difference as @xmath101/[r_{\rm e}(z_{850})+r_{\rm e}(m_{*})],\ ] ] and the average size as @xmath102.\ ] ] we use these two calculations so as to avoid biasing the analysis when the values become very small in either the stellar mass image , or in the @xmath6 band . we plot the relation between these values in figure [ fig : size_cf ] . the horizontal dashed line marks the position of equal size ( @xmath103 ) . the symbols are the same as in figure [ fig : cmd ] , and the average dispersion for the sample is represented by the black square points on the right of the plot . the larger black square points in the figure show the average values and the measured dispersion for the ranges of @xmath104 between zero and one kpc , one and two kpc , up to six kpc . this convention is used for each of the parameter comparisons in sections [ sec : conc_cf ] to [ sec : m20_cf ] . we note that there is a steady progression for increasing average galaxy size with morphology , from the compact ellipticals , early types and early spirals to the late - type spirals , with the peculiars and edge - on galaxies being more randomly distributed . the points are scattered about the line of equality such that nearly half ( 42 percent ) of the galaxies have @xmath105 . there is , therefore , a slight tendency for sizes in masses to be higher , but this does not vary significantly with average galaxy size and type ( see table 2 ) . there is no clear tendency for any particular morphological type to be larger in either the @xmath6 or stellar mass image , which suggests that there is no bias introduced by using measurements in @xmath6 band data for size measurements ( trujillo 2007 ; buitrago et al . table [ tab : size_cf ] . shows that the average difference between the sizes in the stellar mass maps and the @xmath6-band image is essentially zero , demonstrating that the measurements of half - radii in light does not differ significantly from the measurements in the distribution of stellar mass . we also examine the normalised difference in size against ( @xmath16 - @xmath6 ) colour to test whether there is a tendency for bluer galaxies to have larger radii in stellar mass than in light . this is what we might expect to find if star formation dominated the light near the centre of the galaxy . while , on average , colour does not change with size difference , we find that extreme half - radii differences between stellar mass and @xmath6 are mostly in blue systems . and stellar mass . the error bar to the right of the plot shows the average dispersion for the whole sample . the large black squares denote the average values in equally spaced bins of mean size , with error bars showing the dispersions of these values . this convention is used in all of the parameter comparison plots . the symbols plotted here are the same as in figure 12.,width=340,height=340 ] c r r + type & mean size & normalised size difference + & ( kpc ) & + + ce & @xmath106 & @xmath107 + pe & @xmath108 & @xmath109 + e & @xmath110 & @xmath111 + es & @xmath112 & @xmath113 + ls & @xmath114 & @xmath115 + eo & @xmath116 & @xmath117 + pec & @xmath118 & @xmath119 + pm & @xmath120 & @xmath121 + m & @xmath122 & @xmath123 + analogous to the galaxy size comparison in the previous section , we compare the concentration parameter in @xmath6 and stellar mass images , as plotted in figure [ fig : conc_cf ] . the dashed line shows @xmath124 and the black square points represent the average values in equally spaced bins of average concentration . the average concentration , in both @xmath6 and stellar mass , is generally high for the early types , and lower in the late - types and peculiars , with the early - type spirals being spread between low and high values . the compact elliptical galaxies have a low average concentration , due to the light being spread evenly across a galaxy s pixels . the average trend shows that the ratio @xmath125 is slightly lower in the stellar mass maps for the early types compared to the @xmath6 band concentration . we find that in general r@xmath126 is lower and r@xmath54 is higher , compared to those in the @xmath6-band image , although the effect is mostly due to a smaller r@xmath126 although this effects is quite small . it is possible that stellar mass is less centrally concentrated towards the centre of the galaxy in the early - types ( raising r@xmath54 ) , and/or the stellar mass is more diffuse at larger radii than the distribution of light ( lowering r@xmath126 ) . it is also the case that for many galaxies more star formation is seen in the outer regions of the galaxies in the @xmath6-band , but these pixels are less dominant in stellar mass , raising the value of r@xmath126 creating higher values of @xmath10 as seen . band and within stellar mass maps . the error bar to the right of the plot shows the average dispersion for the whole sample . the large black squares denote the average values in equally spaced bins of mean concentration , plus dispersions . the symbols used here are the same as in figure 12.,width=340,height=340 ] before we discuss the comparison of our stellar mass map asymmetry values to the @xmath6-band asymmetries , we note a few things . first , the lowest asymmetry values for the sample are negative in all bands , mainly due to the background correction which for very symmetric galaxies will sometimes be larger than the asymmetry in the galaxy itself . this has the effect , as seen in figure [ fig : asym_cf ] , of skewing the low average asymmetry values differences for the smallest stellar mass measured asymmetries . as can be seen in figure 16 , except for these galaxies with low asymmetry values , there is a clear tendency for galaxies to be more asymmetric in stellar mass than in @xmath6 , this being the case for a large fraction of the non - early type sample . the stellar mass maps also have a greater spread in asymmetry values than @xmath6 , and this can be more clearly seen in [ sec : ac ] . however , we also note that the blue regions of galaxies are not so well traced by the stellar mass maps , and often led to difficulties with the image contrast , making the galaxy features harder to see , as in the case of the late - spiral 19535 ( fig . [ fig : massmap_ls ] ) . further galaxies , such as 38722 ( figure 6 ) , are clearly more asymmetric and lopsided when viewed in the stellar mass band . in this case , it appears that one spiral arm remains while the other disappears . there are several reasons why the asymmetry value in the stellar mass maps is higher than in the images . first , when calculating @xmath11 , noise tends to be magnified due to the process of subtracting images . the simulations discussed in 3.4 show that there is a tendency for the average galaxies to become more asymmetric in stellar mass measurements than in light . this can explain part of this , but we find that the average difference of around @xmath127 in the relative difference is too high to be accounted for solely by these redshift effects . we discuss some of the reasons for why the asymmetries will be higher in the stellar mass maps than in light . when calculating stellar masses for each pixel we assume that the @xmath50 ratio is approximately constant in the surrounding pixels . this should be the case for early type galaxies , especially , due to their uniform stellar populations . to test how much this variance in @xmath50 could be affecting the @xmath11 values , we have measured the @xmath50 pixel variance in a typical early type galaxy ( 30976 , fig . [ fig : massmap_e ] ) . the mean @xmath50 for a pixel in this galaxy is @xmath128 with a typical standard deviation in the surrounding pixels of @xmath129 . although a correction has been applied to minimise this effect ( see 3.3.1 ) , it is not enough to account for the high asymmetry signal in the spiral galaxy stellar mass maps , and thus this asymmetry is likely a real effect due to the nature of the calculation of this parameter , and especially the background , as described in 3.3.1 . and stellar mass . the points and conventions are the same as in figure 12 . note that the asymmetry in the stellar mass is lower than the @xmath25band for only the early types . this is due to the method of measuring the asymmetry , where the sky background is handled in a different way than for imaging , resulting in higher values for later type galaxies ( 4.3.3 ) . , width=340,height=340 ] to further understand why late - type spiral galaxies are more asymmetric in stellar mass than @xmath6 , we have examined these galaxies in detail and present four examples of the late - type sample in both @xmath6 ( fig . [ fig : spirals_z ] ) , and stellar mass maps . two of these galaxies have low asymmetries in both stellar mass and @xmath6 ( 25751 and 21448 ) , and galaxy 21448 in particular , demonstrates the smoothing of spiral arms , which we observed previously in nearby galaxies ( lcm08 ) . however , two of these galaxies have higher asymmetries in stellar mass than in @xmath6 ( 19535 and 34946 ) . object 19535 , especially , shows little relation in stellar mass to the @xmath6 image . the star forming knots and central bulge , which can be seen clearly in @xmath6 , translate to large scale asymmetries in stellar mass . there are several possible explanations for this effect . the star forming regions in 19535 could truly be more massive per pixel than the surrounding disk , due to the high masses and densities of gas and dust required to form such massive star forming regions . this would lead to such regions not being smoothed out in stellar mass , causing higher @xmath7 values . it is also possible that what we have interpreted as star forming knots in the disk in some cases could be minor merger signatures , that is , low - mass galaxies falling into the object itself . there is also a trend with morphological - type with increasing asymmetry such that early types have low asymmetry and late - types higher asymmetry . our findings confirm previous studies who found similar trends ( e.g. conselice 2000 ) . we have also investigated the relation between asymmetry in stellar mass and @xmath130 colour , and examine this after splitting into two redshift bands : @xmath131 and @xmath132 . there is a clear trend for galaxies with a higher degree of asymmetry to be blue , with @xmath133 and @xmath134 . this trend has also been found for asymmetries measured in light ( conselice 2003 ) . and stellar mass . the points and conventions are the same as in figure 12.,width=340,height=340 ] parameter in @xmath6 and stellar mass . the points and conventions are the same as in figure 12 . because of the inverse nature of m@xmath54 , galaxies with positive differences on the y - axis are those with more ` concentrated ' light profiles than in the stellar mass maps.,width=340,height=340 ] the clumpiness values in stellar mass maps have a higher degree of uncertainty , as discussed in section 3.3.3 . overall , we find a trend for clumpiness to be greater in stellar mass than @xmath6 , and this is the case for 91 percent of the sample . however for larger values of average @xmath12 , there is a trend such that @xmath135 . the clumpiness parameter is similar to asymmetry , but picks out small scale features , such as compact star clusters , as opposed to large scale asymmetries . it is , therefore , not surprising that uncertainties in @xmath50 cause a large number of the sample to have @xmath136 , as variations in @xmath50 from pixel to pixel filter through in the calculation of @xmath137 . such variations affect @xmath7 and @xmath137 more than the other parameters , as these calculations involve the subtraction of images , which amplifies the @xmath50 variations and large variations of level within the background . figure [ fig : gini_cf ] shows the comparison between the values of the gini parameter in stellar mass and the @xmath6 band . although the early - types are more strongly clustered around equality , @xmath138 , there is a clear trend for the sample to have higher @xmath70 in stellar mass , with @xmath139 in non - early types within our sample . this comparison of @xmath140 and @xmath141 displays a greater separation between types . the late - type spirals show an approximately even spread in gini values , from 0.45 to 0.93 , but all of these have a higher difference between @xmath6 and stellar mass than the early - types and compact ellipticals . figure [ fig : gini_cf ] reveals that the gini index is higher within the stellar mass maps than in the @xmath6 band , except for early - type galaxies , indicating that most of the stellar mass in later morphological types is contained within fewer pixels within the stellar mass image . this is a significant difference between the early and late - types in our sample . the reason for this is that the bright blue regions of these galaxies vanish as the m / l ratio is inversely proportional to l. this creates effectively a mass distribution with a larger fraction of the mass contained within fewer pixels . for example the bulges become more prominent for the late - type disks while the arms vanish . the bulges in these systems become more prominent and this will rise the value of the gini index . what can often been seen in stellar mass maps of late - types and mergers / peculiars is that these bright outer regions vanish , leaving only the central part of the galaxy including most objects in figure 6 which loose their outer parts and appear almost as early - types . figure 18 also shows that the gini index for the early - types in stellar mass and light are similar , likely because of the small variation in the m / l ratios of the various pixels . we show the differences between the m@xmath54 index in the stellar mass maps and the @xmath6 band in figure [ fig : m20_cf ] . due to the reversed parity in the m@xmath54 values , points which lie above the dashed line in figure [ fig : m20_cf ] have m@xmath142 m@xmath143 , that is , the stellar mass is more more diffuse than the light for most systems . this would change however if we used a smaller radius than the standard definition . overall , m@xmath144 is higher ( less negative ) than m@xmath143 for most of the sample . the m@xmath54 parameter traces the brightest 20 percent of the flux in the galaxy , or the greatest 20 percent of the stellar mass . this parameter is heavily weighted by the spatial distribution of the most luminous , or most massive , pixels , but is normalised to remove any dependence on galaxy size and total flux / stellar mass . m@xmath54 , like @xmath70 , is also not dependent on a fixed , pre - determined centre . therefore , if stellar mass exactly follows light within a galaxy we would expect that m@xmath144 = m@xmath143 . figure [ fig : m20_cf ] shows that , while this is approximately the case for early - type galaxies and early spirals , the same does not apply for peculiars , late - type spirals and edge - on discs . the compact ellipticals especially differ from this trend , having m@xmath144 @xmath145 m@xmath54(@xmath6 ) . this is certainly due to the fact that these compact ellipticals have blue cores which render their stellar mass maps more compact in the centre than for galaxies with redder cores . we examine how galaxies fall in the classic concentration - asymmetry plane ( e.g. , conselice et al . 2000 ) to determine if different galaxy types , as determine visually can be better separated at different wavelengths in this parameter space . we plot the relation between asymmetry and concentration in light for the sample in fig . [ fig : ac_bviz ] and in stellar mass in fig . [ fig : ac_mass ] . over - plotted are the classification criteria of conselice ( 2003 ) and bershady et al . galaxies with @xmath146 are classified as mergers , while those which lie above the top line are early - types . galaxies to the left of the middle line in figure 20 & 21 , are labelled mid - types and those to right of this line are classified as late - types . these criteria for identifying mergers and late - types appear more appropriate for our sample in the @xmath17 and @xmath6 bands , while galaxies are more mixed in these regions in stellar mass , having many more galaxies with high asymmetries , and spirals are not easily distinguished from merging systems . however , the uniqueness of the early - type region is more obvious in stellar mass maps , with most e galaxies in the @xmath17 and @xmath6 band plots falling into the mid - type region ( fig . [ fig : ac_mass ] ) . we find that for our sample , @xmath6 is the best band in which to measure these parameters for separating morphological types from each other . it can also be seen in fig . [ fig : ac_bviz ] that the @xmath11-@xmath10 relation breaks down when viewed in bluer bands , as also shown for nearby galaxies in taylor - mager et al . we find that the scatter in @xmath11-@xmath10 increases for all galaxy types towards bluer wavelengths , the effect is more pronounced for late - type morphologies . for @xmath6 ( top ) and the stellar mass maps ( bottom ) . in @xmath6 ( top ) the dashed lines are the classification criteria of lotz ( 2004 ; equations [ eq : lotz_mergerline ] and [ eq : lotz_normalline ] ) while the solid lines mark our revised criteria in @xmath6 ( equations [ eq : me_mergerline ] and [ eq : me_normalline ] ) . in stellar mass ( bottom ) the lotz criteria are marked by the dashed lines , our @xmath6 criteria are shown by the dot - dashed lines and the solid line shows our separation of early and late types / mergers , in stellar mass.,width=340,height=604 ] ) , which is approximately constant across redshifts ( @xmath31 ) . the dashed line illustrates the fit to the late - type spiral galaxies ( eq . [ eqn : ls_am_z ] ) . the late - type spiral galaxies show a larger @xmath7 with increasingly higher redshifts.,width=340,height=340 ] figure [ fig : ac_mass ] shows the @xmath11-@xmath10 relation for the stellar mass maps . there is a greater range in asymmetry values in stellar mass and the criterion suggested by conselice ( 2003 ) for merging systems , of @xmath55 , even includes a few e and some es galaxies . there is , however , still a clear separation in types between early and peculiar type systems , but the spirals ( especially ls galaxies ) significantly overlap between the two . this makes it difficult to distinguish between morphological types in stellar mass within @xmath147 . however , given the fact that many of these spirals appear to have some kind of merger or formation mode , the stellar mass structure is superior for finding galaxies in active evolution . we plot the relation between @xmath70 and m@xmath54 for light and stellar mass in figures [ fig : gm20_bviz ] and [ fig : gm20_mass ] . the solid lines in fig [ fig : gm20_bviz ] are taken from lotz ( 2008 ) , who used these parameters to classify their sample of normal galaxies and ulirgs . equation [ eq : lotz_mergerline ] describes the line above which lotz classify their sample as mergers , and eq . [ eq : lotz_normalline ] describes the line separating early and late - type galaxies . @xmath148 @xmath149 we again see that the cleanest , in terms of separating morphological types , optical relation is in @xmath6 , although our sample is not best represented by the lotz classifications in gini - m@xmath54 space , with many early - type galaxies falling into the merging region of the plot . this is partially due to the fact that @xmath70 and m@xmath54 are calculated differently here then in lotz , as described in 3.3 and lisker ( 2008 ) . however , we can not rule out that the classifications for this sample may be intrinsically different , and it is worrying that all of the edge - on galaxies fall into the merging region of @xmath70-m@xmath54 . we also use a similar procedure to look at the same rest - frame wavelength as lotz et al . we modify the lotz et al . ( 2008 ) relations for @xmath6 ( top ) and stellar mass ( bottom ) and show this in fig . [ fig : gm20_mass ] . the solid lines in the @xmath6 plot are our revisions to the lotz relations , based on our sample ( equations [ eq : me_mergerline ] for mergers and [ eq : me_normalline ] for normal systems ) , @xmath150 @xmath151 while the dashed lines are the original lotz relations . in the bottom panel of fig . [ fig : gm20_mass ] , the solid line shows our revision of the merger defining line for the stellar mass maps , ( eq . [ eq : mass_mergerline ] ) , @xmath152 while here our relations for @xmath6 and those of lotz are shown as dot - dashed and dashed , respectively . as with the relation for @xmath11(@xmath153)-@xmath10(@xmath153 ) , the @xmath70(@xmath153)-m@xmath144 relation shows the same general trend as its @xmath6 counterpart , but with a larger spread in values . the spirals , again , are not as neatly defined as in @xmath6 , with the es galaxies separated from the early - types and the late - type spirals and edge - on disk galaxies occupying the merging region . however , there is a clear separation in the stellar mass gini vs. @xmath4 which does not exist for the optical light . several of the pe galaxies lie in the merger region , even as defined by the higher stellar mass line ( eq . [ eq : mass_mergerline ] ) . this has also been noted by conselice ( 2007 ) , and is likely due to these objects having multiple nuclei . the @xmath70(@xmath153)-m@xmath144 relation shows a clear early / late - type split , defined by eq . [ eq : mass_mergerline ] , as the sb / sc / irr region , as described by lotz is not appropriate to apply to our sample in stellar mass . we discuss other correlations : asymmetry vs. clumpiness , concentration vs. m@xmath54 , size vs. concentration , and asymmetry vs. m@xmath54 in the appendix a. we evaluate in this appendix other scaling relationship between these non - parametric relationships , including asymmetry - clumpiness , concentration and m@xmath54 , size and concentration , and asymmetry and m@xmath54 . in this section we investigate the changes in our galaxy stellar mass maps with redshifts . we take a quantitative approach here and find that at @xmath0 there is little change in the stellar mass structure with redshift , with the exception of the asymmetries of these galaxies which decreases for some galaxy types at @xmath0 . we plot @xmath7 against redshift ( figure [ fig : am_z ] ) where we can see a general trend for galaxies to become more asymmetric in stellar mass at higher redshifts . we find that not all galaxy types follow this pattern however , with the most notable exception being the ellipticals . as shown in fig . [ fig : am_z ] there is no trend for the ellipticals to become more asymmetric in stellar mass at higher redshifts . we note that while the es remain at approximately the same low asymmetries across the range in redshift , there is a trend for spiral galaxies to become increasingly asymmetric at higher redshifts . we measure this trend by fitting both the ellipticals ( solid line ) and ls ( dashed line ) galaxies , plotted in figure [ fig : am_z ] and quantified in equations [ eqn : e_am_z ] and [ eqn : ls_am_z ] , below . @xmath154 is the best fit for the ellipticals and for the late - type spirals , the best fit is , @xmath155 equation [ eqn : e_am_z ] shows that the early - type galaxies retain approximately constant asymmetries across @xmath156 . equation [ eqn : ls_am_z ] indicates that there is a relation for late - type spirals , such that @xmath7 increases with redshift . we have already discussed in 4.3.3 how the late - type spiral galaxies have stellar mass asymmetries larger than their asymmetries in optical light . since the late - type spirals are galaxies which have arms / disks that are larger / brighter than the bulge component , these galaxies are therefore dominated in terms of their light by their spiral structures and arms . these galaxies show the most diversity and contract between spatial distributions within their m / l and stellar mass maps ( e.g. , fig . [ fig : massmap_ls ] ) . we note that not all of the late - type spirals have high asymmetries in the stellar mass maps , some are as asymmetric or less than in the @xmath6-band ( fig . [ fig : asym_cf ] ) . furthermore , we find that in general the disk galaxies with higher stellar mass map asymmetries have a bluer colour , which is the case at both high and low redshifts . in many of these cases , the structures are lopsided and/or contain outer features that remain after normalising by the m / l map . this interpretation of stellar mass being more localised than the light is consistent with a higher gini index in the stellar mass bands than in the light . some of these asymmetric spirals , as well as many of the peculiars resemble the so - called `` clump clusters '' and clumpy spirals found by elmegreen ( 2005 , 2007 ) , especially in stellar mass maps . their low concentration in mass is also similar to the luminous diffuse objects ( ldos ) found at similar redshifts ( conselice et al . indeed , there are some late - type spirals galaxies that we would have classified differently , perhaps as pec or pm systems , had we performed the classification in stellar mass , rather than in @xmath6 . the clumps in galaxies found by elmegreen are also massive , with typical values in the region @xmath157 ( elmegreen & elmegreen 2005 ) , and reside in young disks at high redshift . it is likely that some of our highly asymmetric late - type spiral galaxies are similar , or in the same class as these clumpy spirals . these features may also be the result of minor merger events . we conduct a pixel by pixel study of the structures of 560 galaxies found in the goods - n field at @xmath0 within stellar mass maps and hubble space telescope acs @xmath2 wavebands . we measure stellar masses for each pixel of each galaxy image from our @xmath2 images by fitting to stellar population models . we use these values to construct stellar mass maps for our sample and compare morphologies in @xmath2 and stellar mass . our major findings and results include : i. we construct stellar mass maps and mass - to - light ratio maps for each galaxy and we present examples of each morphological type , in the @xmath6 band , stellar mass maps , and @xmath98 in figures [ fig : massmap_ce ] to [ fig : massmap_m ] . we find that some compact elliptical ( ce ) galaxies have blue cores , and more complicated internal structures than in either @xmath6 or stellar mass would reveal on their own . many of the peculiar ellipticals ( pe ) also have blue cores and we assert that these objects may have merged in their recent histories . the early and late - type spirals display more varied patterns . we find that for some of these galaxies , structures seen in light ( e.g. spiral arms ) are smoothed out in stellar mass . however , this effect does not hold true for all galaxies , nor for all features seen in light . stellar mass and ( @xmath50 ) maps of the ` peculiar ' galaxies are complicated and vary across the sample . although it is more difficult to make out features in stellar mass for galaxies that have very blue colours , structures not seen in light are revealed . we compare the half - light radius , @xmath5 , in @xmath6 with half - mass radii for our sample as a function of morphological type . we find no systematic tendency for any particular morphological type to have larger @xmath5 in stellar mass than in the @xmath6 band , and thus conclude that there is no bias introduced by measuring galaxy sizes in @xmath6 . we find a clear tendency for many galaxies to be more asymmetric in stellar mass than in @xmath6 . we also find that morphology correlates with asymmetries , with early - types having low @xmath7 , and late - types higher values of @xmath7 . we find a relation between colour and @xmath7 such that bluer galaxies have higher stellar mass asymmetry differences between optical light and stellar mass maps . the late - type spiral galaxies in the sample have higher @xmath7 than would be expected from their asymmetries in @xmath6 . we discuss possible causes of this effect , including regions of enhanced star formation also possessing higher stellar masses , and the evolution of spirals . we note that these highly asymmetric spirals resemble the clumpy disks of elmegreen ( 2007 ) , and conselice et al . ( 2004 ) and are experiencing either minor merging activity , or bulge formation through accretion of disk material . we find that the gini index in stellar mass is higher than in the z - band ( @xmath139 ) for all galaxies except the early - types , indicating that most of the stellar mass in later morphological types is contained within fewer pixels . late - type and peculiar morphologies show a trend for m@xmath144 @xmath158 m@xmath143 , suggesting that the brightest @xmath159 percent of the brightest pixels is not necessarily where the greatest @xmath159 percent of the stellar mass is located . v. we investigate the relations between several combinations of the @xmath9 parameters in @xmath2 and stellar mass , and compare our results to previous morphological studies of this type ( e.g. conselice , rajgor & myers 2008 ; lotz 2004 ; 2008 ) . we find that @xmath6 is the most appropriate photometric band to utilise for a @xmath0 sample of galaxies with all morphologies , although stellar mass maps are better at distinguishing active galaxies from passive ones and is more a physical measure of structure . we furthermore compare our sample classifications in @xmath70-m@xmath54 to those of lotz ( 2008 ) and find that the lotz criteria do not best describe our sample . we revise the lotz ( 2008 ) criteria to best fit our sample in the @xmath6-band and within the stellar mass maps , and find that early - types , late - types and mergers can be separated . however , the edge - on disk galaxies remain problematic and can not be distinguished from mergers in @xmath140-m@xmath143 . we find that @xmath141-m@xmath144 can be used to broadly separate early from late - type galaxies , but the criteria can not distinguish between late - type / edge - on disks from peculiar / merger systems . we find a relationship between @xmath5 and @xmath10 ( see appendix a ) for early - type galaxies in both @xmath6 and stellar mass . in each case we find that galaxies with higher concentrations have larger radii , and this relation is steeper in @xmath6 than stellar mass . we also investigate asymmetry versus m@xmath54 and find that these parameters display a similar relation to @xmath147 in @xmath17 and @xmath6 . in stellar mass , @xmath7-m@xmath144 shows a tighter relation than @xmath160 , with a clear separation between early and late - type systems . thus , we conclude that , for all parameters , late - type spiral galaxies overlap with with pec / pm / m systems which may be ultimately taken from the same subset of galaxies . structural studies in stellar mass do however track both minor and major merging events , and can be used to find galaxies in active galaxy evolution modes . we thank the goods team for making their data public , and stfc for a studentship towards supporting this work and the leverhulme trust for support . 99 abraham , r.g . , et al . 1996 , mnras , 279 , 47 abraham , r.g . , et al . 1999 , mnras , 303 , 641 abraham , r.g . , van den bergh , s. , nair , p. 2003 , aj , 588 , 218 beauvais , c. , bothun , g. 1999 , apjs , 125 , 99 bershady , m.a . , jangren , j.a . , & conselice , c.j . 2000 , aj , 119 , 2645 bertone , s. , conselice , c.j . 2009 , mnras , 396 , 2345 blanton , m.r . & roweis , s. 2007 , aj , 133 , 734 bothun , g.d . , 1986 , aj , 91 , 507 bruzual , g. & charlot , s. 2003 , mnras , 344 , 1000 buitrago , f. , trujillo , i. , conselice , c.j . , bouwens , r.j . , dickinson , m. , yan , h. 2008 , apj , 687 , 61l cassata , p. , et al . 2010 , apj , 714l , 79 carrasco , e.r . , conselice , c.j . , trujillo , i. 2010 , mnras , 405 , 2253 chabrier , g. 2003 , apj , 586 , 133l conselice , c.j . , 1997 , pasp , 109 , 1251 conselice , c.j . , bershady , m.a . , & jangren , a. 2000 , apj , 529 , 886 conselice , c.j . , gallagher , j.s . , wyse , r.f.g . 2002 , aj , 123 , 2246 conselice , c.j . 2003 , apjs , 147 , 1 conselice , c.j . et al . , 2003 , aj , 126 , 1183 conselice , c.j . , et al . 2004 , apj , 600 , 139l conselice , c.j . , blackburn , j. a. , papovich , c. , 2005 , apj , 620 , 564 conselice , c.j . 2006a , apj , 638 , 686 conselice , c.j . 2006b , mnras , 373 , 1389 conselice , c.j . , et al . 2007 , mnras , 381 , 962 conselice , c.j . , rajgor , s. & myers , r. 2008 , mnras , 386 , 909 ( paper i ) conselice , c.j . , yang , c. , bluck , a.f.l . 2009 , mnras , 394 , 1956 ( paper iii ) conselice , c.j . , arnold , j. 2009 , mnras , 397 , 208 elmegreen , d.m . , elmegreen , b.g . , rubin , d.s . , schaffer , m.a . 2005 , apj , 631 , 85 elmegreen , d.m . , elmegreen , b.g . , ravindranath , s. , coe , d.a . 2007 , apj , 658 , 763 ferguson , h.c . , et al . 2004 , apj , 600 , 107l giavalisco , m. , et al . 2004 , apj , 600 , 103l graham , a.w . , driver , s.p . , petrosian , v. , conselice , c.j . , bershady , m.a . , crawford , s.m . , goto , t. 2005 , aj , 130 , 1535 hernandez - toledo , h.m . , avila - reese , v. , conselice , c.j . , puerari , i. 2005 , 129 , 682 lanyon - foster , m.m . , conselice , c.j . & merrifield , m.r . 2007 , mnras , 380 , 571 lisker , t. 2009 , apjs , 179 , 319 lotz , j. , primack , j. , madau , p. 2004 , aj , 128 , 163 lotz , j.m . , et al . 2008 , apj , 672 , 177 maraston , c. 2005 , mnras , 362 , 799 mobasher , b. , et al . 2004 , apj , 600 , 167l oke , j.b . & sandage , a. 1968 , apj , 154 , 21 petrosian , v. 1976 , apj , 209 , 1l riciardelli , e. , trujillo , i. , buitrago , f. , conselice , c.j . , 2010 , mnras , 406 , 230 simard , l. , et al . 2002 , apjs , 142 , 1 stanford , s.a . , et al . 2004 , aj , 127 , 131 strauss , m.a . , et al . 2002 , aj , 124 , 1810 taylor - mager , v.a . , conselice , c.j . , windhorst , r.a . , jansen , r.a . 2007 , apj , 659 , 162 trujillo , i. , conselice , c.j . , bundy , k. , cooper , m.c . , eisenhardt , p. , ellis , r.s . 2007 , mnras , 382 , 109 weinzirl , t. , et al . 2011 , apj , 743 , 87 welikala , n. , et al . 2011 , arxiv : 1112.2657 welikala , n. , connolly , a.j . , hopkins , a.m. , scranton , r. , conti , a. 2008 , apj , 677 , 970 windhorst , r. , et al . 2002 , apjs , 143 , 113 wirth , g.d . , 2004 , aj , 127 , 3121 in this appendix we describe several additional scaling relationships between structural and morphological parameters , not discussed in the body of the paper , for our galaxy sample . many of the relations we see here confirm the conclusions we reached earlier . furthermore , we list these for a complete overview of how structure correlates in the hubble space telescope acs @xmath2 bands as well as in stellar mass maps . we postulate that the many small scale asymmetries , measured by the clumpiness parameter , must approximate to the asymmetry of the galaxy as a whole . figures [ fig : as_bviz ] and [ fig : as_mass ] show the relation between @xmath11 and @xmath12 for @xmath18 , @xmath16 , @xmath17 , @xmath6 and stellar mass , respectively . the horizontal solid line in each plot marks the @xmath161 position and the dashed line in each plot illustrates where @xmath162 . the solid diagonal lines in each plot are the fits to the e , es and ls galaxies in each band , which were forced to coincide at @xmath163 . these fits are shown below in equations [ eqn : as_b]- [ eqn : as_mass ] , for each band respectively . . the horizontal dashed line marks the position of @xmath164 , which we derive as a criterion for merger candidates in stellar mass.,width=283,height=283 ] @xmath165 @xmath166 @xmath167 @xmath168 @xmath169 it is evident that the relations are mostly steeper than @xmath162 , although the fit in stellar mass comes closest to this equality . the @xmath11(@xmath153)-@xmath12(@xmath153 ) relation thus is useful for identifying merger candidates . the criterion of @xmath11(@xmath153 ) @xmath170 is too low for the stellar mass maps , instead we suggest that galaxies which lie in the region above @xmath11(@xmath153 ) @xmath171 ( horizontal dashed line , fig . [ fig : as_mass ] ) and underneath the @xmath11(@xmath153)-@xmath12(@xmath153 ) fit , be classified as mergers , and this could be adapted for use to calculate merger fractions across a range of redshifts . the trend noted for the parameters to be less reliable in the bluer bands also holds here ( fig [ fig : as_bviz ] ) . we note that the criteria @xmath172 and @xmath173 are appropriate for identifying merger candidates , and calculating merger fractions . the fits to @xmath11-@xmath12 for @xmath17 , @xmath6 and stellar mass are very similar , despite the much larger scatter in @xmath11(@xmath153 ) . for the stellar mass maps . the solid line shows our fit to the e and es galaxies in stellar mass ( eq . [ eqn : cm20_mass ] ) . the solid line is this relation plus @xmath174 . the points are the same as in figure a1.,width=283,height=283 ] in this section we investigate the relationship between m@xmath54 and @xmath10 in @xmath18 , @xmath16 , @xmath17 , @xmath6 and stellar mass ( figures [ fig : cm20_bviz ] and [ fig : cm20_mass ] ) . the solid lines in each plot show the fits to the e and es galaxies in the sample . the forms of these fits are shown below in equations [ eqn : cm20_b ] to [ eqn : cm20_mass ] . @xmath175 @xmath176 @xmath177 @xmath178 @xmath179 m@xmath54 is anti - correlated with @xmath10 , such that highly concentrated galaxies have lower m@xmath54 values . the general trend for concentration to increase as m@xmath54 decreases holds for all bands , although the scatter increases at higher values of m@xmath54 and lower @xmath10 . the relation has a larger dispersion in the @xmath18 and @xmath16 bands , but this is more ordered than the random scatter seen in previous relations ( fig . [ fig : cm20_bviz ] , top panels ) . the relation is distinct in both @xmath17 and @xmath6 , with the early - type galaxies occupying the high @xmath10 , low m@xmath54 region , the es s having higher m@xmath54 values for their concentration and late - types generally having low @xmath10 and high m@xmath54 . this trend is also seen in the stellar mass maps ( fig . [ fig : cm20_mass ] ) , but with greater separation between the morphological types than in @xmath17 and @xmath6 . the early - types also have high concentration and low m@xmath54 , although the spirals are more randomly distributed , and do not occupy any one region of the plot . the late / pec / pm / m galaxies have higher m@xmath54 values for their concentrations . the dashed line in fig . [ fig : cm20_mass ] shows the fit plus @xmath174 and we postulate that galaxies falling above this criterion can be picked out as merger candidates . in this section we investigate the relation between galaxy size , as measured by the half - light radius , @xmath5 , and concentration , in @xmath6 and stellar mass . the top panel in fig . [ fig : rec_zm ] shows the relation in @xmath6 and the bottom panel in stellar mass . we find that the early - types follow a linear relationship , and we have fit this relation for the e galaxies in @xmath6 ( eq . [ eqn : rec_z ] ) and stellar mass ( eq . [ eqn : rec_mass ] ) . and concentration for @xmath6-band images ( top ) and stellar mass maps ( bottom ) . the solid lines show the fits to the e galaxies in each band ( equations [ eqn : rec_z ] and [ eqn : rec_mass ] ) . the points are the same as in figure a1.,width=340,height=680 ] @xmath180 @xmath181 the relation in @xmath6 is steeper than in stellar mass , and both trends show that as @xmath5 increases , @xmath10 also increases . for normal galaxies , @xmath10 is a tracer of stellar mass , such that lower @xmath10 corresponds to less massive galaxies ( conselice 2003 ) . in both @xmath6 and stellar mass we see that more concentrated galaxies have larger radii , and this effect is more pronounced in @xmath6 than in stellar mass . no such relation is found for the late - types and pec / pm / m systems . conselice & arnold ( 2009 ) examine galaxies at @xmath182 in the hubble ultra deep field ( hudf ) finding that early - types are smaller for their stellar mass at higher redshifts . they conclude that those galaxies that follow the @xmath183 relation are in a relaxed state . here we investigate the relation between asymmetry and m@xmath54 in the @xmath18 , @xmath16 , @xmath17 , @xmath6 bands ( fig . [ fig : am20_bviz ] ) and stellar mass ( fig . [ fig : am20_mass ] ) . as m@xmath54 is analogous to @xmath10 , we would expect to see a similar relation between @xmath184 and @xmath147 . in fig . [ fig : am20_bviz ] we see that in the @xmath2 bands we indeed find a similar relation in @xmath184 to @xmath147 , and could be used alongside @xmath147 for classification purposes . for the stellar mass maps . the solid diagonal line marks our division between early and late - types / mergers.,width=283,height=283 ] in stellar mass , however , the relation between @xmath184 appears tighter than @xmath185 , such that galaxies with higher asymmetries have lower values of @xmath4 . there is a clear separation between early and late - type galaxies in stellar mass , and we have included an approximate divide in this plot ( solid line in fig . [ fig : am20_mass ] ) , the form of which is shown below in equation [ eq : am20_mass_div ] . @xmath186 although this divide is very successful in separating early - types from pec / pm / m systems , the spirals are spread over these two regions , again implying that measuring structural parameters in stellar mass does not distinguish between morphologies as well , although it traces physical processes better . however , it is possible that viewing the stellar mass distribution of galaxies reveals different physical processes to those we expect to see in light . we have already discussed the possibility that some highly asymmetric late - type spirals in our sample could be in the process of forming bulges . it is also possible that studies in stellar mass could provide clues as to the formation mechanism involved . the asymmetric late - type spirals , for example object 19535 ( figs . [ fig : massmap_ls ] , [ fig : spirals_z ] ) , could be undergoing outside - in formation , perhaps through the accretion of smaller satellites . we could , therefore , be picking out systems experiencing minor merging events . at the start of this paper we describe how we classified each of our galaxies into one of nine type ( 3.2 ) . in this section , we described , based on our stellar mass map analysis , the properties of each of these galaxy types in terms of their optical and stellar mass structures , and what these imply about the evolution of different classes of galaxies . while galaxy classification in terms of overall visual morphology is limited ( conselice 2006b ) , it is still useful for connecting galaxies at different redshifts , and for understanding how major nearby galaxy types such as disks and ellipticals are formed . below we give details on the properties of each of the nine galaxy type that we have examined within goods . compact ellipticals are typically found in the late - type area of the c - a optical light diagram , but are within the early - type region of the c - a stellar mass diagram . the clumpiness values for these systems are also zero in all wavelengths and stellar mass due to the fact that these systems are so small this parameter can not be reliably measured . the concentrations of these systems are always quite low , with @xmath187 , at the resolution limit where we can measure these values , suggesting that they are not similar to giant ellipticals ( e.g. , conselice 2003 ) . some interesting features appear when we compare the quantitative properties of ces as measured in optical light to those in stellar mass maps . one of these is that for all ces , the size measured in the stellar mass maps are always larger than for those measured in the reddest band , @xmath6 . the asymmetry measured in stellar mass maps is also quite low , and lower than in the , perhaps due to the fact that the asymmetric light is brought to a lower level within the stellar mass images compared to the optical light . the m@xmath54 parameter is also quite different between the stellar mass maps and the , such that m@xmath54 is higher ( less negative ) in the . this implies that the light distribution is more concentrated in the stellar mass maps , likely due to the blue coloured core . first , we find that just based on colours , there is a wide diversity at a given redshift ( see figs . 12 and 13 ) - an indication that at higher redshifts there is a range of elliptical galaxy star formation histories , something that has been recognised for some time ( e.g. , stanford et al . 2004 ; conselice et al . 2007 ) . simply based on the c - a diagram with optical light ( fig . 20 ) it appears that most of the early - types are found in the region of c - a space classifiable as mid - types . this is commonly seen for high redshift ellipticals which appear more asymmetric than their @xmath188 counterparts ( e.g. , conselice et al . ellipticals do have generally lower gini and m@xmath54 indices , however there is still some spread in their values . the stellar mass maps reveal that ellipticals appear to become more ` diffuse ' within stellar mass maps . however the sizes are either very similar in stellar mass maps or even slightly smaller , than in the corresponding @xmath6-band image . the gini and the m@xmath54 parameters do not change significantly between the stellar mass maps and the @xmath6-band image . however , we find that some of the ellipticals are slightly more asymmetric in stellar mass maps than in the @xmath25band , although many of these systems tend to be bluer galaxies . the ellipticals also tend to become less concentrated in stellar mass compared with the @xmath6 band ( fig . [ fig : conc_cf ] ) . finally , the elliptical galaxies are one of the few types which do not change significantly in terms of their gini and m@xmath54 values between stellar mass and the @xmath6 , and this results in these systems having a well defined location in the stellar mass gini / m@xmath54 diagram ( fig . [ fig : gm20_mass ] ) . the peculiar ellipticals are a galaxy type which has been recognised to exist up to at least @xmath189 before ( e.g. , conselice et al . these peculiar ellipticals appear elliptical in overall morphology , but have internal structure suggesting that they have recently undergone some type of assembly . these galaxies are not uncommon , and are more often seen than the compact elliptical type . these galaxies are more asymmetric and less concentrated than the es - with all having concentrations , @xmath190 . they are quite distinct from the ellipticals in the ca plane and are bluer by @xmath191 magnitudes from the giant ellipticals . these blue features can be readily seen within the m / l maps for these systems ( fig . [ fig : massmap_pe ] ) . these galaxies often appear more elliptical in the stellar mass maps once their m / l variations have been taken into account . another difference between the pes and the es is that these systems have half - light ( or half - mass ) radii which are nearly always bigger in the stellar mass maps than in the @xmath6 band . this is a sign for star formation activity in the cores of these galaxies , since the bluer material in the core is less prominent in the stellar mass maps , giving them a larger effective radius . this can also be seen in the fact that the asymmetry in the stellar mass maps is generally higher than in the @xmath6 band for these systems , which is another indication that they are undergoing some form of star formation . the gini values for these systems are higher in the stellar mass band than in @xmath6 , which demonstrates that the stellar mass distribution is more concentrated in bright ( but non - central ) pixels than the light , as the light is less equally distributed than the stellar mass . the location of the pes is similar to that of the es in the gini / m@xmath54 plane . the spiral galaxies are amongst the most interesting for this study given that they often contain two major stellar population types segregated spatially . traditionally this is seen as an older stellar population making up the bulge or centre , and the spiral arms consisting of younger stellar populations . we find that the early - type spirals , those whose apparent bulge is brighter / larger than its apparent disk , are quite blue in colour . this difference is very obvious when examining the m / l ratio maps for these galaxies ( fig . [ fig : massmap_es ] ) , although the differences are not as obvious as with the late - type spirals . nevertheless , these systems often appear to have a blue outer m / l ratio and a redder inner one , although there are examples where this is not the case . the sizes of these galaxies does not change significantly between the stellar mass maps and the @xmath6 band images . there are however some differences between the stellar mass maps and the @xmath6 band images within the various parameters . the most significant of these is that the asymmetry parameter is higher in the stellar mass maps than in the @xmath6 band imaging as discussed in detail in 4.3.3 . the gini parameter is also higher within the stellar mass maps than within the @xmath6 by roughly 10 percent . this implies that the stellar mass is distributed in fewer pixels than in the light for spirals . the distribution of m@xmath54 values is similar between the light and stellar mass , suggesting that the spatial distribution is similar . one notable feature of the spiral galaxies in general is that they have a much larger range of asymmetry values within the stellar mass images . in fact , their values are often as high as the peculiar galaxies , suggesting that within stellar mass maps these parameters can not easily distinguish between mergers and spirals with mixed stellar populations . the peculiars , pre - mergers and mergers all have similar patterns in the cas space in both the stellar mass maps and within the @xmath6 diagrams . these are the most asymmetric and bluest galaxies in our sample , and are likely in some phase of a merger , as has been described earlier in e.g. , conselice et al . ( 2003 , 2008 ) . these peculiars ( a term we use here for all three of these types ) do not change in size much between the stellar mass and the @xmath6 image . however , the asymmetry in the stellar mass band is higher than for the @xmath6 band . this is likely due to the stellar mass within these galaxies being more concentrated in fewer locations . the m@xmath54 also shows that the light is less concentrated in than the stellar mass , showing that the stellar mass is more distributed spatially than in light , suggestive of a merging systems . in fact , these peculiars have the highest asymmetries , m@xmath54 , and gini indices .	galaxy structure and morphology is nearly always studied using the light originating from stars , however ideally one is interested in measuring structure using the stellar mass distribution . not only does stellar mass trace out the underlying distribution of matter it also minimises the effects of star formation and dust on the appearance and structure of a galaxy . we present in this paper a study of the stellar mass distributions and structures of galaxies at @xmath0 as found within the goods fields . we use pixel by pixel k - corrections to construct stellar mass and mass - to - light ratio maps of 560 galaxies of known morphology at magnitudes @xmath1 . we measure structural and size parameters using these stellar mass maps , as well as on acs @xmath2 band imaging . this includes investigating the structural @xmath3gini-@xmath4 parameters and half - light radius ( @xmath5 ) for each galaxy . we further identify and examine unusual galaxy types with this method , including compact and peculiar ellipticals , and peculiar galaxies in some mode of formation . we compare structural parameters and half - light radii in the acs @xmath6-band and stellar mass maps , finding no systematic bias introduced by measuring galaxy sizes in @xmath6 . we furthermore investigate relations between structural parameters in the acs @xmath2 bands and stellar mass maps , and compare our result to previous morphological studies . combinations of various parameters in stellar mass generally reveal clear separations between early and late type morphologies , but can not easily distinguish between star formation and dynamically disturbed systems . we also show that while ellipticals and early - type spirals have fairly constant cas values at @xmath0 we find a tendency for late - type spiral and peculiar morphological types to have a higher @xmath7 at higher redshift . we argue that this , and the large fraction of peculiars that appear spiral - like in stellar mass maps , are possible evidence for either an active bulge formation in some late - type disks at @xmath0 or the presence of minor merger events . [ firstpage ] galaxies : structure , morphology , classification , evolution
i would like to thank f.e.low for his collaboration in deriving eq . many fruitful discussions i would like to thank u.g . meiner , n.kaiser , b.holstein , and u.b.vankolk . for critical help with the manuscript i would like to thank m.distler and m.pavan . s. weinberg , transactions of the n.y . academy of science series ii 38 ( 1977 ) , 185 ( i.i . rabi festschrift ) and article in the proceedings of the workshop on chiral dynamics : theory and experiment , springer - verlag , july 1995 , a.m. bernstein and b. holstein editors . meiner and s. steininger , phys . b419 ( 1998 ) 403 . n. fettes , u.g . meiner , and s. steininger , hep - ph/9803266 ( to be published in nucl . phys . ) . meiner , plenary talk at the `` quark - lepton physics 1997 '' , osaka , japan , may 1997 , hep - ph/9706367 . d. sigg et . al . , phys . lett . 75 ( 1995 ) 3245 , a609 ( 1996 ) 269 , a617 ( 1997 ) 526 . a. badertscher , in proceedings of the workshop on chiral dynamics , mainz . sept 1 - 5 , 1997 , a.m. bernstein , d. drechsel , and th . walcher , editors , springer - verlag , in press . , and private communication .	it is demonstrated that there is a dynamic isospin breaking effect in the near threshold @xmath0 reaction due to the mass difference of the up and down quarks , which also causes isospin breaking in the @xmath1 system . the photopion reaction is affected through final state @xmath1 interactions ( formally implemented by unitarity and time reversal invariance ) . it is also demonstrated that the near threshold @xmath2 reaction is a practical reaction to measure isospin breaking in the @xmath1 system , which was first predicted by weinberg about 20 years ago but has never been experimentally tested . since the discovery of a large mass difference between the up and down quarks ( @xmath3 ) there has been considerable interest in the possibility of observing dynamical isospin breaking in the pion - nucleon system @xcite . the theoretical consensus is that the magnitude of isospin breaking is not @xmath4 , but instead is @xmath5 @xcite , where @xmath6 0.2 gev . in this paper it will be shown for the first time that the @xmath7 reaction is an excellent candidate to measure dynamic isospin violations due to the up , down quark mass difference . weinberg first showed that there is an isospin violating effect in the s wave @xmath1 scattering length @xmath8 due to the up , down quark mass difference @xcite . this predicted effect , which occurs for @xmath9 scattering or charge exchange but not for @xmath10 scattering , has more recently been calculated by chiral perturbation theory ( chpt ) @xcite . the predicted magnitude of this effect is the same ( to within a factor of @xmath11 in @xmath12 scattering and charge exchange reactions . however , since the magnitude of @xmath13 is small , the relative magnitude of the isospin violating term is @xmath14 30 % @xcite . by contrast , for charge exchange where the isospin conserving amplitude is larger , the relative isospin violation is estimated to be @xmath14 2 to 3 % @xcite . in this paper the ( dynamic ) isospin breaking effect of the up and down quark mass difference is shown to be present in the near threshold @xmath15 n @xmath16n reaction . this is in addition to the well known ( static ) isospin breaking effect due to the threshold difference between the @xmath17 and @xmath18 channels . of this @xmath14 2/3 is due to the coulomb effect and @xmath14 1/3 is caused by the up , down quark mass difference . the separation of the two threshold energies leads to the prediction of a unitary cusp in the @xmath19 reaction near the @xmath20 threshold due to the two step @xmath21 reaction @xcite . the magnitude of this unitary cusp is due to the large ratio of the electric dipole multipole amplitudes , @xmath22 @xcite . the first derivations of the unitary cusp @xcite used a k matrix approach to calculate the effect of charge exchange in the final state . due to their single scattering approximation , these calculations violate unitarity as will be discussed below . the chiral perturbation theory ( chpt ) calculations are basically isospin conserving but the biggest isospin non - conserving effect due to the pion mass difference ( which is almost purely electromagnetic in origin ) is inserted by hand @xcite . these calculations also violate unitarity due to their truncation at the one loop ( single rescattering ) level . there has been recent progress in performing chpt calculations in @xmath1 scattering which take the isospin violations due to the coulomb interaction into account @xcite . the approximations made in the k matrix @xcite and published chpt papers @xcite can be overcome by using a 3 channel s matrix approach in which unitarity and time reversal invariance are satisfied ( a preliminary version of this work has been presented previously @xcite ) . this has the advantage that both static and dynamic isospin breaking , and final state multiple scattering ( to all orders ) , are taken into account . the s matrix for the 3 open channels @xmath23 can be written as : @xmath24 where @xmath25 represent elastic @xmath26 , @xmath27 , @xmath20 scattering with phase shifts @xmath28 , and @xmath29 respectively , @xmath30 represents the @xmath31 charge exchange amplitude , @xmath32 is a real number , and @xmath33 represents the inelasticity due to charge exchange . for convenience , the off diagonal matrix elements for the photo - production of the @xmath27 and @xmath20 channels are written as @xmath34 and @xmath35 , where @xmath36 and @xmath37 are proportional to the multipole amplitudes . for the important case of production of s wave pions these are the @xmath38 transverse ( longitudinal ) multipoles . although not explicitly written here , the s matrix elements are for a fixed value of w , the total cm energy , and the quantum numbers l and j , the @xmath39 orbital and total angular momenta . the notation of using 0 ( c ) for the neutral ( charged ) channel is conveniently generalized to neutron targets where the three channels are @xmath40 , and @xmath41 . time reversal invariance requires that the s matrix be symmetric ( @xmath42 ) and unitarity requires that @xmath43 . the form of the 2x2 @xmath1 portion of the s matrix has been chosen to be separately unitary and time reversal invariant . eq . [ eq1 ] is symmetric by construction . applying the unitary constraint and assuming the weakness of the electromagnetic interaction by dropping terms of order e@xmath44 , one obtains @xcite : @xmath45 \\ m'_{c } & = e^{i(\delta_{\gamma}+\delta_{c } ) } [ a'_{c } \cos \frac{\phi}{2 } + i a'_{0 } \sin \frac{\phi}{2 } ] \end{array}\label{eq2}\ ] ] where @xmath46 and @xmath47 are smoothly varying , real functions , of the cm energy , and can be identified as proportional to the multipole matrix elements @xmath48(for @xmath17 ) and @xmath49 ( for @xmath50 ) in the absence of the final state @xmath1 interaction . these equations show the connection between electromagnetic pion production and @xmath1 interactions . note that the phase shifts @xmath28 , and @xmath29 appear as @xmath51 in photopion reactions compared to @xmath52 in elastic scattering because the @xmath53 interactions take place in the final ( initial ) state only . eqs . [ eq1 ] and [ eq2 ] are valid for both photo- and electro - production , i.e. for both real and virtual photons . in general @xmath54 , and @xmath47 are functions of @xmath55 ( the invariant four momentum transfer ) as well as w , l , and j. to order @xmath56 the @xmath1 sector parameters @xmath57 , and @xmath32 are functions of w , l , and j only . in this treatment terms of order @xmath56 have been neglected with the exception of @xmath58 which is small compared to both @xmath59 and @xmath29 ( except at the @xmath60 threshold ) , and the mass difference between the charged and neutral pions which is put in by hand . this is the largest @xmath61 effect found in @xmath1 scattering @xcite and in photoproduction @xcite . the effect of the @xmath61 terms have been worked out , although not presented here . in general they produce a small coupling between the electromagnetic and @xmath1 sectors , so that e.g. the 2x2 @xmath1 sector of the s matrix is no longer separately unitary . when the experimental situation reaches a sufficient level of accuracy these corrections should then be implemented . eq . [ eq2 ] is a generalization of the final state interaction theorem of fermi and watson @xcite who first pointed out the connection between photo - pion production and @xmath1 scattering . in their derivation , as in this one , time reversal and unitarity ( to order @xmath56 ) were assumed . however they made the additional assumption that if one could neglect @xmath56 ( and higher order terms ) isospin would be conserved ( this was before the time of quarks ) . this reduces the dimensionality of the s matrix to 2x2 which gives the simple solution @xmath62 where i=1/2,3/2 is the isospin of the @xmath1 system . if isospin were conserved , so that the two thresholds are degenerate , eq . [ eq2 ] reduces to the fermi - watson theorem . for the near threshold region in which the cm pion momentum q @xmath63 , the s wave phase shifts are @xmath64 q/3 = a_{\pi^{0}p } q , \delta_{c } = [ a_{3}+2a_{1}]q/3= a_{\pi^{+}n } q$ ] , and @xmath65 q/3 = a_{cex}q$ ] where @xmath66 is the s wave @xmath1 scattering lengths in the isospin 2i = 1 , 3 , or designated charge states . however isospin conservation is badly violated in the threshold region due to the threshold energy difference between the @xmath67 and @xmath68 channels and the additional dynamic isospin violating effect due to the up and down quark mass difference . this generalization of the fermi - watson theorem removes the approximation of isospin conservation . note the interesting feature that below the @xmath20 threshold there are only two open channels and eq . [ eq2 ] reduces to the form of the fermi - watson theorem , i.e. @xmath36 = real number @xmath69 or , equivalently , that @xmath70 . it is of interest to show the connection between the s matrix formulation which takes the final state scattering into account to all orders and the original k matrix derivations @xcite which only took a single final state scattering into account . this single scattering limit can be taken by expanding @xmath71 , and neglecting second order terms like @xmath72 , thus obtaining @xmath73 . by examing the region below the @xmath20 threshold where @xmath74 is continued as @xmath75 , it is observed that @xmath76 this means that the single scattering approximation does not obey unitarity , although in this particular case the numerical effects are not large . the chpt calculations to one loop have a similar unitarity problem as was discussed by the chpt authors @xcite . the connection between @xmath77 to the multipole amplitudes @xmath78 is @xmath79 . note that the multipoles ( m ) have the dimensions of length , while the s matrix elements ( m ) are dimensionless . the s wave photoproduction cross section is @xmath80 charge exchange reaction cross section is @xmath81 from which one obtains @xmath82 @xcite . defining @xmath83 one can then write eq . [ eq2 ] in terms of multipole amplitudes as : @xmath84 \\ m_{c } & = e^{i(\delta_{\gamma}+\delta_{c } ) } [ a_{c } \cos \frac{\phi}{2 } + i a_{0 } q_{0 } \| f_{cex } \|/\cos \frac{\phi}{2 } ] \end{array}\label{eq3}\ ] ] in the threshold region eqs . [ eq2 ] and [ eq3 ] lead to a unitary cusp . in this case , @xmath85 , and @xmath86 for the transverse multipole , or @xmath87 for the longitudinal multipole . below the @xmath20 threshold @xmath74 is continued as @xmath88 . thus the contribution of the two step @xmath89 reaction is in @xmath90 below ( above ) the @xmath20 threshold . the magnitude of the unitary cusp , @xmath91 , is defined as the coefficient of @xmath74 : @xmath92 the value of @xmath91 can be calculated from the experimental value of @xmath93 @xcite . assuming isospin is conserved @xmath94 @xcite . for @xmath95 the chpt prediction of @xmath96 @xcite is used . this is in agreement with the preliminary value of @xmath97 from a recent experiment @xcite . combining the two experimental ( theoretical ) numbers gives @xmath98 @xmath99 @xcite . although it has not been previously noticed , the same physical mechanism that causes the isospin breaking in @xmath1 scattering and charge exchange also enters into @xmath91 . this follows from unitarity and time reversal ( eqs . [ eq2 ] and [ eq3 ] ) , combined with the prediction of isospin breaking for the s wave scattering lengths for charge exchange reactions @xmath100 , that there is an isospin violating contribution to @xmath91 , @xmath101 in the second line it has been assumed that there is no quark mass effect on @xmath102 . it is estimated that @xmath103 2 to 3% @xcite . a better theoretical calculation of this effect should be performed . the electric dipole amplitude @xmath104 with its unitary cusp is presented in figs . [ fig1 ] and [ fig2 ] . the recent mainz / taps @xcite and saskatoon @xcite results for @xmath105 are presented in fig . [ fig1 ] , where only the statistical errors are shown . the small deviations between the data sets suggest the magnitude of the systematic errors . considering this , the agreement between the calculated curves and the data is satisfactory . the curves are the chpt calculation with three empirical low energy parameters used in fitting the data @xcite , and a unitary fit to the data @xcite using eq . [ eq3 ] which has @xmath106 @xcite and a linear function of photon energy for @xmath107 with two parameters which were fit to the data . for this case @xmath108 @xcite and the effect of the two step charge exchange reaction is dramatic in the real ( imaginary ) part below(above ) the @xmath20 threshold . approximately 1 mev above the @xmath20 threshold the @xmath109 . the negative sign in @xmath110 makes @xmath111 have the opposite sign from @xmath112 . it is interesting to note that both the power counting rules of chpt and the constraints of unitarity lead to pion rescattering in the final state as a critical dynamical ingredient . although the unitary and chpt curves both agree with the data for @xmath105 there is an important difference between them . the value of @xmath91 = 2.78 @xcite calculated for the @xmath17 reaction is smaller then the one calculated using the separately predicted values of @xmath113 and @xmath114 of @xmath115 . this difference can be clearly seen in fig . [ fig2 ] . the reason for this discrepancy , which was discussed by the chpt authors @xcite , is due to the fact that the chpt calculation is carried out to one loop which is not sufficient for @xmath111 . this is a general feature of chiral perturbation theory in which the imaginary part of the amplitude is not calculated as accurately as the real part , and thus unitarity is only approximately satisfied at a given order @xcite . as will be discussed below the difference between the unitary and one loop chpt values of @xmath91 can be observed in future experiments in which @xmath111 will be measured directly . to accurately measure the magnitude of the unitary cusp and to exploit the connection between electromagnetic pion production and low energy @xmath1 interactions , one must measure @xmath111 . in photoproduction this requires experiments with polarized beams and/or targets @xcite . for the sake of brevity the formulas connecting the cross sections and polarization observables to the multipoles @xcite will not be quoted here . to briefly demonstrate the power of polarized photo - pion experiments , two asymmetries are shown in fig . [ fig3 ] : @xmath116 for linearly polarized photons with an unpolarized target ; and t for unpolarized photons but with a target polarized normal to the reaction plane . the results presented in fig . [ fig3 ] use the p wave predictions of chiral perturbation theory @xcite and the unitary fit to @xmath117 discussed above @xcite . @xmath116 is primarily sensitive to the p wave multipoles and since the unitary fit has essentially the same p wave multipoles as chpt , the curves for chpt and the unitary fit are almost identical . by contrast , t is sensitive to a linear combination of p wave multipoles times @xmath111 , and shows its rapid rise above the @xmath20 threshold . for t , the large difference in @xmath111 between the unitary fit @xcite and chpt to one loop @xcite should be straightforward to distinguish experimentally . a proposed experiment with tagged photons using an active , polarized proton target at mainz estimates that @xmath91 can be measured to @xmath118 to 2% @xcite . similar results could be obtained using a laser backscattering source @xcite or small angle electron scattering with polarized , internal , targets in a storage ring facility @xcite . there are several possible strategies to test isospin conservation from measurements in the near threshold @xmath119 reaction . a definitive demonstration of this effect would involve precision measurements of the real and imaginary parts of @xmath117 for the unitary cusp contributions to the @xmath120 and @xmath17 reactions . although the predicted effect of @xmath14 2 to 3% in @xmath91 is larger than the usual @xmath141% effect expected on the basis of electromagnetic effects , it represents a serious experimental challenge to accomplish this goal , particularly since free neutron targets are not available to make this measurement . a simpler strategy would be to perform a precision measurement of the unitary cusp for the @xmath121 reaction , extract the value of @xmath91 , and compare the result of with the value obtained using unitarity and isospin conservation quoted above . an equivalent strategy is to obtain @xmath122 from the measured value of @xmath91 and compare it to the accurately measured value of @xmath123 @xcite , obtained from the width of the 1s state in pionic hydrogen , to see if there is any deviation from the pure isospin prediction that these values are equal and opposite @xcite . it is important to note that there is a psi proposal @xcite to improve the pionic hydrogen measurement so that the error in @xmath123 will be reduced to @xmath14 1% . it is therefore of interest to discuss the information that might be obtained from measurements of the s wave @xmath1 scattering lengths , @xmath8 , from photopion reactions with the values that have been obtained using conventional pion beams . the values for @xmath8 for several physical channels are presented in table [ tab1 ] ( the appropriate isospin relations @xcite were used when required ) . the most accurate and direct measurements come from pionic hydrogen experiments @xcite . in order to test isospin symmetry the psi group has also measured pionic deuterium in order to get at the @xmath124 scattering length @xcite . unfortunately the two body corrections in deuterium are large and add significantly to the error which makes the isospin conservation test uncertain at the required level of accuracy , so that these values are not quoted here . the results of the two most recent empirical analyses of the @xmath1 data for pion kinetic energies greater than @xmath14 30 mev , extrapolated to threshold , are also presented @xcite in table [ tab1 ] . in the said study the scattering lengths come from a dispersive analysis , the errors are hard to ascertain and are not quoted @xcite . the chiral perturbation theory results @xcite are consistent with experiment . if isospin symmetry is exact there are two independent scattering lengths in the i = 1/2 , 3/2 states , and six possible physical @xmath125n elastic scattering and charge exchange reactions , so that there are many possible tests if three or more channels can be measured precisely . this is precisely the gap that can be filled by measurement of the final state @xmath125n interactions in photopion production , e.g. @xmath27 or @xmath126 in the @xmath127 reaction . .s - wave @xmath1 scattering lengths @xmath8 for several channels in units of @xmath128 ( n = n or p ) [ cols="^,^,^,^,^,^ " , ] another test of isospin conservation has been made for medium energy @xmath125n scattering ( pion kinetic energy from 30 to 100 mev ) @xcite . in both cases the scattering amplitudes in the i = 1/2 and 3/2 states were obtained from the data for @xmath129 elastic scattering . from these , the prediction for the @xmath130 charge exchange amplitude made on the assumption of isospin conservation was compared to the empirical charge exchange amplitude . an isospin violation at the @xmath131 level has been found by both analyses @xcite , primarily in the s wave amplitude . these analyses depend on the quality of the data and on the accuracy of the coulomb corrections , which have been criticized as being inconsistent with the strong interaction calculations that were employed @xcite . if isospin is indeed violated at the 7% level , it is not clear how to relate this to the isospin breaking predictions for the s wave scattering length @xcite which is applicable at lower energies . what is required is an extension of the isospin breaking calculations to medium energies . a test of the existing predictions requires experiments at lower energies which would measure the s wave scattering lengths . in addition to the measurement of @xmath111 above the @xmath20 threshold discussed above , one could contemplate the more difficult measurement when only the @xmath27 channel is open . as was discussed previously , from a measurement of both the @xmath132 and @xmath133 parts of @xmath117 ( or @xmath134 ) one obtains @xmath59 , for which there is a predicted @xmath135 isospin breaking effect . however , because of the small expected size of @xmath136 in the energy region below the @xmath20 threshold , this is a very difficult task . it may be at the limits of feasibility using either a laser backscattering source , where an intense photon flux is concentrated in a small energy interval @xcite , or with internal target , small angle scattering with polarized internal targets @xcite . in conclusion , the connection between the @xmath137 reaction and @xmath138 scattering has been presented in a rigorous , model - independent way , which is unitary and time reversal invariant , and where the isospin breaking due to the threshold difference between the @xmath68 and @xmath67 channels , and the mass difference between the up and down quarks , is taken into account . this leads to a predicted unitary cusp due to the two step @xmath139 charge exchange reaction . the magnitude of the unitary cusp is given by @xmath140 . this unitary cusp has been recently observed in photoproduction experiments @xcite . it has been shown for the first time that there is a dynamical isospin breaking effect in the value of @xmath91 , due to the mass difference of the up and down quarks , in electromagnetic pion production . this is linked by unitarity and time reversal invariance to a predicted quark mass effect in @xmath141 scattering and pion charge exchange @xcite . at present there are accurate measurements for @xmath142 and @xmath143 from pionic atoms performed at psi @xcite . to check isospin conservation requires at least one more precision measurement in another charge channel . this is more readily performed in electromagnetic meson production for two reasons : 1 ) in order to accurately measure the s wave scattering length , it is important to work at very low energies ( e.g. @xmath144 10 mev ) at which @xmath1 experiments are hard to perform since the low energy charged pions decay and also since @xmath60 beams can not be made at any energy ; and 2 ) in electromagnetic pion production one can access charge states that can not be reached with conventional pion beams ( this is important for isospin checks ) . it is shown that photoproduction experiments with polarized targets can lead to a precise measurement of the small but interesting isospin violating effects . in addition to the polarization observables in photoproduction discussed above , similar information can be obtained in threshold @xmath60 electroproduction from the combination of the transverse- longitudinal ( tl and tl ) structure functions @xcite . this possibility will be discussed in a future publication . finally we stress that the important program of precisely measuring the @xmath1 and photopion reactions that are required to test the predicted isospin violation , is difficult but feasible .
in most usual laboratory and/or astrophysical conditions chemical reactions of type @xmath0 are well described by the kinetic approach @xcite . if @xmath1 , @xmath2 are the concentrations of species @xmath3 and @xmath4 , and @xmath5 , the problem is fixed by the relative concentrations @xmath6 suppose some method is given for picking particles of types @xmath3 and @xmath4 while sampling the system . concentrations @xmath7 and @xmath8 are then the relative probabilities of getting particles of the corresponding species at time @xmath9 . the reaction rates are typically given by the inverse times of free - flight , or velocity/(mean free path ) ratios : if @xmath10 is the average velocity of type-@xmath11 particles , @xmath12 where @xmath13 and @xmath14 are the corresponding reaction cross - sections . the kinetic picture underlying such definitions is well known @xcite : in eq.([reactionrate11 ] ) , for example , @xmath15 is the effective cylindric volume presented by particle @xmath16 to particle @xmath3 per second . particle @xmath16 will consequently meet @xmath17 particles @xmath3 per second . the relative concentrations as functions of time are then described by the master ( or gain / loss ) equations @xmath18 variation in the abundance of species @xmath3 is the abundance of species @xmath4 times the rate of @xmath4-to-@xmath3 transformation ( which represents the gain ) minus the @xmath3 abundance times the rate of its disappearance ( the loss ) . let us introduce the notations @xmath19 and @xmath20 . situations are not unusual in which both @xmath21 and @xmath22 are very nearly constant ( see section [ sec : comments ] ) . in that case , it is possible to obtain general analytical solutions for the above master equations . the problem reduces to solving the differential equation @xmath23 with constant coefficients . by their very meanings , @xmath24 and @xmath25 . evolution will cease when @xmath26 , which suggests two candidate equilibrium values : @xmath27 . once equilibrium is attained , the backward reaction is as important as the forward reaction and eq.([eq : reaction ] ) is , of course , better written with a two - sided arrow , @xmath28 the @xmath29 term in the right - hand side of ( [ masternp31 ] ) is present only if @xmath30 . in that case the solution is @xmath31\right\ } , \label{eq : solutionxc}\ ] ] with @xmath32 an integration constant whose determination will later provide our main result . as lim@xmath33 , this solution tends indeed to one of the above candidate equilibria : @xmath34 the solution for @xmath35 is obtained by simply exchanging parameters @xmath21 and @xmath22 . equilibria probabilities are related by @xmath36 only to provide some intuitive guidance , plots for toy models @xmath37 and @xmath38 are shown in figures [ reactionac ] and [ reactionac2 ] . and @xmath39 . upper line : starting with no @xmath3 ; lower line : starting with pure @xmath3 . equilibrium is attained with @xmath40 . _ , title="fig:",width=283,height=151 ] and @xmath41 . equilibrium is attained with @xmath42 . _ , title="fig:",width=283,height=151 ] an analogous result relates @xmath43 to @xmath44 . contact with the usual equilibrium approach @xcite can be made through a few simple considerations . in the equilibrium classical ( nonrelativisitic , non quantal ) case , particle @xmath4 ( for example ) will have concentration @xmath45 , where @xmath46 counts the values taken by `` internal '' degrees of freedom ( spin , isospin , etc ) , @xmath47 is the de broglie thermal wavelength of particle @xmath4 ( @xmath48 ) and @xmath49 is its chemical potential . the equilibrium condition is @xmath50 or , in terms of the fugacities @xmath51 , @xmath52 . direct comparison with the above result leads to @xmath53^{3/2 } = \sqrt{\frac{v_{a } v_{b}}{v_{c } v_{d}}}\,\,\ , \frac{\sigma_{_{ab\rightarrow cd } } } { \sigma_{_{cd\rightarrow ab } } } \,\ , .\ ] ] if we use the equipartition formula @xmath54 , @xmath53^{5/4 } = \frac{\sigma_{_{ab\rightarrow cd } } } { \sigma_{_{cd\rightarrow ab } } } \,\ , .\ ] ] we see that the cross - sections have , in this case , just to account for kinematic factors . if we take navely @xmath55 , @xmath56 , @xmath57 , @xmath58 , the expected trivial equilibrium requirement follows : @xmath59 . situation @xmath60 , which must be considered separately , would turn up in the peculiar case @xmath61 : the volume spanned by @xmath62 per unit time , as seen by @xmath4 , equals the volume spanned by @xmath16 per unit time , as seen by @xmath3 . the solution of eq.([masternp31 ] ) would , in that case , be @xmath63 with , naturally enough , the probabilities tending to equilibrium at @xmath64 . an example would be an `` elastic '' reaction of type @xmath65 with the same cross - section @xmath66 in both sides . the crossed reaction @xmath67 would be accounted for by the general case , as @xmath68 , equilibrium being given by the condition @xmath69 . there are two main approaches to evolving systems . we have above used the first : time evolution is described by a continuous curve of type @xmath70 , solution of some differential equation . in the second , evolution is described by the successive iterations of a mapping @xcite . the state is known at each step , as if the `` time '' parameter of the system were defined only at discrete values . we can go from the first approach to the second by taking the intersections leading to a poincar map . this approach supposes a characteristic time the time of a unit step . if a continuous description can be shown to be the interpolation of a discrete mapping @xcite , a clear notion of characteristic time obtains . there is , however , a strong requirement : that interpolation must preserve the notion of iteration all along . this requirement is encapsulated in the so - called semigroup conditions @xcite . for a function @xmath71 describing the dynamical flow of a system , these conditions are @xmath72 = f^{<t'>}[f^{<t>}(x ) ] = f^{<t+t'>}(x ) \ ; \ ; ; \label{i.1 } \\ f^{<0>}(x ) = x \ ; . \label{i.2}\end{gathered}\ ] ] a sufficient condition for that is that the solution have the form @xmath73 , \label{contiteratebriggs}\ ] ] for some function @xmath74 , its inverse @xmath75 and a constant @xmath76 . this would mean that @xmath74 solves the schrder functional equation @xmath77= c \ ; f(x).\ ] ] this can be translated into the additive form @xmath78 = c ' + f(x)$ ] by taking @xmath79 $ ] what matters is that the semigroup conditions be respected . let us now notice that the integration constant @xmath32 in ( [ eq : solutionxc ] ) can be obtained by simply taking the inverse function at @xmath80 . that solution assumes then the form @xmath81\right\ } . \label{eq : solutionxc2}\ ] ] this is actually the continuum form of an iterate discrete mapping , and fulfills the semigroup requirement . in more detail : introduce the notations @xmath82 for the composition of functions @xmath83 and @xmath84 , @xmath85 for the @xmath86-th iterate of @xmath83 and @xmath87 for the function inverse to @xmath83 . then , with the functions @xmath88 expression ( [ eq : solutionxc2 ] ) is in effect the continuum version of @xmath89 the quantity @xmath90 is the one - step time in the iteration and indicates the time interval in which the reaction process does make significant progress . solution ( [ eq : specialcase ] ) for the special @xmath60 case has already been written in iterative form , with @xmath91 and characteristic time @xmath92 , half the time of free flight . the assumption used above that @xmath21 and @xmath22 are constant actually mean that time @xmath93 is short in comparison with any other `` macroscopic '' time - scale involved . such a `` macroscopic '' time can , for instance , be the time in which temperature and/or volume of the system change appreciably under the influence of some external agent . for reactions of cosmological interest , as those involved in primordial nucleosynthesis , it is the inverse rate of expansion , or inverse hubble function . the quantities ( velocities and cross sections ) appearing in eq.([eq : equilibriumx ] ) will depend on such large - scale time . the first criterion for the validity of the above results is , consequently , that @xmath93 be very short in comparison to those times . in order to use the equilibrium formulae , it would be necessary that equilibrium be attained in not too many steps . it is usual to take the average time of free flight @xmath94 as an order - of - magnitude indication of the lapse necessary for thermalization to be established . it gives a rough measure of the time between two `` hits '' in the reaction . this parameter , however , turns up under conditions quite different from those supposed above @xcite . it is conceived for an arrangement by which particles of type @xmath3 impinge with constant velocity @xmath95 upon a medium formed by particles of type @xmath16 with constant number density @xmath96 . with ( constant ) cross section @xmath97 , the mean free path of an @xmath3 particle in the medium will be @xmath98 = @xmath99 , and the corresponding time of free flight , @xmath100 . the number of unscattered ( that is , keeping the same momentum direction ) particles @xmath3 will then be given by @xmath101 this is a pure - loss equation , quite the same as that for radiative decay . eventual reproduction of particles @xmath3 with the original momentum by other scatterings is neglected , so that there is no gain . the solution is , of course , the radiative decay formula @xmath102 this corresponds , up to normalizations , to the special solution ( [ eq : specialcase ] ) . expression ( [ eq : chartime ] ) , @xmath103 coming from gain - loss considerations , takes also the backward reaction into account and provides , in principle at least , a far better measure of the reaction characteristic time . a deep discussion of the conditions which are supposed to be satisfied can be found in r. balescu , _ statistical dynamics _ , imperial college press , london , 1997 . the chemists approach is described in j.e . leffler and e. grunwald , _ rates and equilibria of organic reactions _ , dover , new york , 1989 , where a very clear exposition is made of the `` cage effect '' and the consequent diffusive regime , which will not be of our concern here . r. aldrovandi , _ special matrices of mathematical physics _ , world scientific , singapore , 2001 . see chap.2 for the two dynamical pictures and , for the relationship between iterates and schrder s equation , 15.2 .	the analytic solution for the kinetic description of binary reactions can be seen as the continuum version of a basic discrete iterate mapping . this fact allows a clear definition of the reaction characteristic time which takes the backward effect into account .
* context and problem statement . * many polynomial systems or systems of laurent polynomials arising in applications do not have a dense monomial structure ( _ e.g _ multi - homogeneous systems , fewnomials , systems invariant under the action of a linear group , ) . the development of toric geometry during the 70s/80s has led to toric ( or sparse ) elimination theory @xcite , a framework designed to study and exploit algorithmically these monomial structures . central objects in toric geometry are _ semigroup algebras _ ( also called toric rings ) . if @xmath7 is an affine semigroup ( see def . [ def : affinesgp ] ) , then the semigroup algebra @xmath8 $ ] is the set of finite sums @xmath9 , where @xmath10 is a formal symbol , @xmath11 is a field , @xmath12 and @xmath13 . semigroup algebras are isomorphic to subalgebras of @xmath14 $ ] generated by a finite subset of monomials . our motivation is to propose fast algorithms to solve symbolically systems whose support lie in one of the following classes of semigroups : semigroups constructed from the points with integer coordinates in a normal lattice polytope @xmath1 ( in that case , the algorithms we propose are well - suited for _ unmixed _ systems : the newton polytopes of the input polynomials are all equal to @xmath2 ) or semigroups generated by a scattered set of monomials ( fewnomial systems ) . * main results . * given a @xmath3-dim . system of laurent polynomials @xmath15 and a finite subset @xmath16 such that each polynomial belongs to the subalgebra generated by @xmath17 , we associate to @xmath18 two affine semigroups : @xmath19 generated by @xmath18 and @xmath20 generated by @xmath21 . under the assumption that @xmath22 contains zero but no nonzero pairs @xmath23 s.t . @xmath24 , our solving strategy proceeds by combining a sparse variant in the homogeneous algebra @xmath25 $ ] of the matrixf5 algorithm and a sparse variant in @xmath26 $ ] of the fglm algorithm . we define a notion of _ sparse grbner basis _ [ def : sgb ] ) that is computed by the sparse - matrixf5 algorithm if we know a bound on its maximal degree ( this maximal degree is called the _ witness degree _ of the system ) . an important feature of sparse gbs is that their definition depends only on the ambient semigroup algebra and not on an embedding in a polynomial algebra . in this sense , they differ conceptually from sagbi bases , even though the sparse - fglm algorithm has similarities with the sagbi - fglm algorithm proposed in @xcite . in the special case @xmath27 , sparse grbner bases in @xmath28 $ ] are classical grbner bases , and sparse - fglm is the usual fglm algorithm . at the end of the solving process , we obtain a rational parametrisation of the form @xmath29 where @xmath30 $ ] is a univariate polynomial , and for all @xmath31 , @xmath32 is a rational function . consequently , the solutions of the input sparse system can be expressed in terms of the roots of the univariate polynomial @xmath33 by inverting a monomial map . the next main result addresses the question of the complexity of this solving process when @xmath18 is given as the set @xmath34 , where @xmath1 is a lattice polytope of dimension @xmath35 . it turns out that the complexities of sparse - matrixf5 and sparse - fglm algorithms depend mainly on intrinsic combinatorial properties of @xmath2 : * the normalized volume @xmath36 ; * the castelnuovo - mumford regularity @xmath37)=n+1-\ell$ ] where @xmath38 is the smallest integer such that the intersection of @xmath39 with the interior of @xmath40 is nonempty ; * the ehrhart polynomial @xmath41 which equals the cardinality of @xmath42 for @xmath43 . we use as indicator of the complexity the _ witness degree _ which bounds the maximal `` sparse degree '' ( corresponding to an @xmath44-grading on @xmath45 $ ] ) in a reduced sparse grbner basis . more precisely , we obtain the following complexity estimates : [ theo : mainres ] let @xmath1 be a normal lattice polytope of dimension @xmath35 with one vertex at @xmath46 , @xmath47 be a sequence of positive integers and @xmath48 be a regular sequence of laurent polynomials in @xmath14^n$ ] , such that the support of @xmath49 is included in @xmath50 . then a sparse gb of the ideal @xmath51 $ ] can be computed within @xmath52 arithmetic operations in @xmath11 , where @xmath53 is a feasible exponent for the matrix multiplication and @xmath54)+1+\sum_{j=1}^n(d_j-1)$ ] . moreover , if @xmath55 is a simple vertex of @xmath2 ( _ i.e. _ a vertex which is the intersection of @xmath35 facets ) , then the sparse - fglm algorithm executes at most @xmath56 arithmetic operations in @xmath11 . direct consequences of these formulas allow us to derive new complexity bounds for solving regular multi - homogeneous systems . we show that the witness degree of a regular system of @xmath35 multi - homogeneous polynomials of multi - degree @xmath57 w.r.t . blocks of variables of sizes @xmath58 ( with @xmath59 ) is bounded by @xmath60 ( which generalizes the bound @xmath4 in the bilinear case @xcite ) . we also propose a variant of frberg s conjecture for sparse systems and a notion of semi - regularity , which yield complexity estimates for solving sparse overdetermined systems . we have implemented in c a prototype of the sparse - matrixf5 algorithm , that runs several times faster than the original @xmath0 algorithm in the fgb software . for instance , we report speed - up ratios greater than 100 for instances of overdetermined bihomogeneous systems . the implementation also works well for fewnomial systems ( although this case is not covered by our complexity analysis ) . * related works . * computational aspects of toric geometry and grbner bases are investigated in @xcite . in particular , ( * ? ? ? * subroutine 11.18 ) gives an algorithm to compute syzygies of monomials in toric rings , which is an important routine for critical - pairs based algorithms . other approaches have been designed to take advantage of the sparse structure in grbner bases computations . for instance , the slim grbner bases in @xcite describes strategies to avoid increasing the number of monomials during computations . this approach improves practical computations , but does not lead to new asymptotic complexity bounds for classes of sparse systems . the sparse structure and the connection with toric geometry have also been incorporated to the theory of resultants , and a vast literature has been written on this topic , see _ e.g. _ @xcite . in particular , mixed monomials structures are well - understood in this context . although we do not know how to extend the algorithms proposed in this paper to mixed structures , grbner - type algorithms enjoy the property of extending without any modification to the overdetermined case . * perspectives . * our approach is for the moment limited to _ unmixed systems _ : all input polynomials have to lie in the same semigroup algebra . a possible extension of this work would be the generalization to mixed systems ( where the algorithms would depend on the newton polytope of each of the polynomials of the system ) . some results seem to indicate that such a generalization may be possible : for instance , under genericity assumptions , mixed monomial bases of quotient algebras are explicitly described in @xcite . also , a bound on the witness degree and the complexity analysis is for the moment restricted to the polytopal case . merging the approach in this paper with a buchberger s type approach such as ( * ? ? ? * algo . 11.17 ) could lead to a termination criterion of the sparse - matrixf5 algorithm in the non - regular cases and for positive dimensional systems . finally , finding complexity bounds which explain the efficiency of the sparse grbner bases approach for fewnomial systems ( see table [ table : fewnomials ] ) remains an open problem . * organisation of the paper . * we recall in section [ sec : background ] the background material on semigroup algebras and convex geometry that will be used throughout this paper . section [ sec : sgb ] introduces sparse grbner bases and describes a general solving process for sparse systems . the main algorithms are described in section [ sec : algos ] and their complexities are analyzed in section [ sec : complexity ] . finally , we describe in section [ sec : applis ] some results that are direct consequences of this new framework and experimental results in section [ sec : expe ] . * acknowledgements . * we are grateful to kaie kubjas , guillaume moroz and bernd sturmfels for helpful discussions and for pointing out important references . this work was partly done while the second author was supported and hosted by the max planck institute for mathematics ( bonn , germany ) . this work was partly supported by the hpac grant of the french national research agency ( hpac anr-11-bs02 - 013 ) . in this paper , the basic algebraic objects corresponding to monomials in classical polynomial rings are _ affine semigroups_. we always consider them embedded in @xmath39 . we refer the reader to @xcite for a more detailed presentation of this background material . first , we describe the main notations that will be used throughout the paper : [ def : affinesgp ] an _ affine semigroup _ @xmath61 is a finitely - generated additive subsemigroup of @xmath39 for some @xmath62 containing @xmath46 and no nonzero invertible element ( _ i.e. _ for all @xmath63 ) . any affine semigroup has a unique minimal set of generators , called the _ hilbert basis _ of @xmath61 and denoted by @xmath64 . let @xmath65 denote the smallest subgroup of @xmath39 containing @xmath61 . then @xmath61 is called _ normal _ if @xmath66 . for a field @xmath11 , we let @xmath8 $ ] denote the associated _ semigroup algebra _ of finite formal sums @xmath67 where @xmath12 . an element @xmath68 $ ] is called a _ monomial_. we use the letter @xmath18 to denote a finite subset of @xmath39 such that @xmath69 and the semigroup @xmath22 generated by @xmath18 contains no nonzero invertible element . also , we let @xmath70 denote the affine semigroup generated by @xmath71 . the semigroup algebra @xmath25 $ ] is _ homogeneous _ ( _ i.e. _ @xmath44-graded and generated by degree @xmath72 elements ) : the degree of a monomial @xmath73 is @xmath74 . the vector space of homogeneous elements of degree @xmath74 in @xmath25 $ ] is denoted by @xmath25_d$ ] . depending on the articles on this topic , the condition `` @xmath61 contains no invertible element '' is not always included in the definition of an affine semigroup . however , this is a necessary condition for the algorithms we propose in this paper . also , the term `` hilbert basis '' is sometimes reserved for affine semigroups of the form @xmath75 where @xmath76 is a rational cone ( see _ e.g. _ ( * ? ? ? * prop . 7.15 ) and the discussion after this statement ) . we always assume implicitely that @xmath77 is a full rank lattice ( this does not lose any generality since this case can be reached by embedding @xmath78 in a lower dimensional @xmath79 ) . note that @xmath80 $ ] is the classical polynomial ring @xmath81 $ ] . semigroup algebras are integral domains ( * ? ? ? 7.4 ) of krull dimension @xmath35 and play an important role in toric geometry : they are precisely the coordinate rings of _ affine toric varieties_. the normality of the semigroup @xmath61 is an important property which implies that @xmath8 $ ] is cohen - macaulay by a theorem by hochster @xcite . an important feature of normal affine semigroups is that they can be represented by the intersection of @xmath39 with a pointed rational polyhedral cone ( also called _ strongly convex rational polyhedral cone _ * sec 1.1 ) ) . a _ cone _ @xmath82 is a convex subset of @xmath83 stable by multiplication by @xmath84 , the set of non - negative real numbers . the dimension @xmath85 of a cone @xmath76 is the dimension of the linear subspace spanned by @xmath76 . a cone is called _ pointed _ if it does not contain any line . a pointed cone of dimension @xmath72 is called a ray . a ray is called _ rational _ if it contains a point in @xmath39 . a _ rational polyhedral cone _ is the convex hull of a finite number of rational rays . pointed rational polyhedral cones will be abbreviated _ prpc_. we shall use prpcs in section [ sec : sgb ] to define admissible monomial orderings in semigroup algebras . we now recall the definition of _ simplicial affine semigroups _ , for which we will be able to derive tight complexity bounds for the sparse - fglm algorithm ( section [ sec : complexity ] ) . [ def : simplicial ] an affine semigroup @xmath86 is called simplicial if the convex hull of @xmath87 is a simplicial prpc , _ i.e. _ the convex hull of @xmath35 linearly independant rays . another important family of objects are _ projective toric varieties_. their homogeneous coordinate rings are associated to a lattice polytope , which we shall assume to be normal in order to ensure that the coordinate ring is cohen - macaulay . as in the classical case , homogeneity is a central concept to analyze the complexity of grbner bases algorithms . all lattice polytopes will be assumed to be full dimensional . a _ lattice polytope _ @xmath1 is the convex hull of a finite number of points in @xmath39 . its _ normalized volume _ , _ i.e. _ @xmath88 times its euclidean volume , is denoted by @xmath36 . to a lattice polytope @xmath1 is associated an affine semigroup @xmath89 generated by @xmath90 . the polytope @xmath2 is called _ normal _ if @xmath91 is a normal semigroup . the associated semigroup algebra is called a _ polytopal algebra _ and abbreviated @xmath92 $ ] . if @xmath1 is a lattice polytope containing @xmath55 as a vertex , then @xmath92=k[s_{{\mathscr p}\cap{\mathbb{z}}^n}^{(h)}]$ ] ( def . [ def : affinesgp ] ) . moreover , if @xmath2 is normal , then so is @xmath93 ( * ? ? ? also , note that if @xmath94 is a translation of @xmath2 , then the homogeneous algebras @xmath92 $ ] and @xmath95 $ ] are isomorphic . consequently , we shall assume w.l.o.g . in the sequel that one of the vertices of @xmath2 is the origin , so that @xmath96 verifies the assumptions of def . [ def : affinesgp ] . we also introduce a few more notations for lattice polytopes : the number of lattice points in a polytope @xmath1 ( _ i.e. _ the cardinality of @xmath97 ) is denoted by @xmath98 . the _ minkowsky sum _ of two lattice polytopes @xmath99 is the lattice polytope @xmath100 . for all @xmath101 we write @xmath40 for the minkowski sum @xmath102 with @xmath38 summands . for @xmath62 , we let @xmath103 denote the _ standard simplex _ , namely the convex hull of @xmath55 and of the points @xmath104 whose entries are zero except for the @xmath105th coefficient which is equal to @xmath72 . for @xmath106,@xmath107 we write @xmath108 for the lattice polytope whose points are @xmath109 next , we recall several useful classical properties of polytopal algebras . we refer to ( * ? ? ? 12 ) for a detailed presentation of the connections between ehrhart theory and computational commutative algebra . [ prop : hspolytope ] let @xmath110 be a lattice polytope . for @xmath74 , we let @xmath111 $ ] denote the _ ehrhart polynomial _ of @xmath2 , _ i.e. _ @xmath112 . also , let @xmath113 $ ] denote the generating series @xmath114 then the hilbert series of the polytopal algebra @xmath92 $ ] , namely @xmath115}(t)=\sum_{d\in{\mathbb{n}}}\dim_k ( k[{\mathscr p}]_d)t^d\ ] ] is equal to @xmath116 and there exists a polynomial @xmath117 $ ] with non - negative coefficients such that @xmath118 the fact that the map @xmath119 is polynomial is a classical result by ehrhart @xcite . the second statement @xmath120}$ ] follows from the definition of @xmath92 $ ] . the last statement is stanley s non - negativity theorem ( * ? ? ? we let @xmath121)$ ] denote the _ castelnuovo - mumford regularity _ of @xmath92 $ ] . the castenuovo - mumford regularity of a graded module is an important measure of its `` complexity '' : it is related to the degrees where its local cohomology modules vanish . we refer to ( * ? ? ? 15 ) for a more detailed presentation . the following classical proposition relates the regularity with a combinatorial property of the polytope @xmath2 and with the degree of the numerator of @xmath116 : [ prop : regpolytope ] let @xmath2 be a normal lattice polytope . the regularity @xmath121)$ ] equals @xmath122 , where @xmath38 is the smallest integer such that @xmath123 contains an integer point in its interior . moreover , with the same notations as in proposition [ prop : hspolytope ] , @xmath124)$ ] . the first claim follows from ( * ? ? ? 5.4 ) . to prove the second claim , we use the partial fraction expansion of @xmath116 which is of the form @xmath125 with @xmath126 . then we obtain the equality @xmath127 , and hence @xmath128 is the smallest positive integer such that @xmath129 . the ehrhart - macdonald reciprocity @xcite concludes the proof . in this section , we show that classical grbner bases algorithms extend to the context of semigroup algebras . first , we need to extend the notion of admissible monomial ordering and of grbner bases . we recall that the monomials of a semigroup algebra @xmath8 $ ] are the elements @xmath130 for @xmath13 . [ def : sgb ] let @xmath61 be an affine semigroup . a total ordering on the monomials of @xmath8 $ ] is called _ admissible _ if * it is compatible with the internal law of @xmath61 : for any @xmath131 , @xmath132 ; * for any @xmath133 , @xmath134 . for a fixed admissible ordering @xmath135 and for any element @xmath136 $ ] , we let @xmath137 denote its leading monomial . similarly , for any ideal @xmath138 $ ] , @xmath139 denotes the ideal generated by @xmath140 . a finite subset @xmath141 is called a _ sparse grbner basis _ ( abbreviated sgb ) of @xmath142 with respect to @xmath135 if the set @xmath143 generates @xmath139 in @xmath8 $ ] . note that admissible orderings exist for any semigroup algebra : the convex hull of a semigroup @xmath144 is a prpc @xmath145 ( this is a consequence of the fact that there is no nonconstant invertible monomial in @xmath8 $ ] ) . now one can pick @xmath35 independant linear forms @xmath146 with integer coefficients in the dual cone @xmath147 , and set @xmath148 if and only if the vector @xmath149 is smaller than @xmath150 for a classical admissible ordering on @xmath151 . note that the assumption that @xmath8 $ ] contains no nonconstant invertible monomial is a necessary and sufficient condition for the existence of an admissible ordering . we describe now an algorithmic framework to solve sparse systems of laurent polynomials . let @xmath152 be a finite subset verifying the assumptions of definition [ def : affinesgp ] , and @xmath153 $ ] be laurent polynomials such that the supports of the @xmath49 are included in @xmath154 . note that translating @xmath18 amounts to multiplying the laurent polynomials by laurent monomials : this does not change the set of solutions of the system in the torus @xmath155 . assuming that the system @xmath15 has finitely - many solutions in @xmath156 , we proceed as follows : 1 . homogenize @xmath157 via def .- prop . [ def : hom ] ( note that the homogenization depends on the choice of the ( not necessarily minimal ) generating set @xmath18 ; 2 . compute a sparse grbner basis w.r.t . a graded ordering of the homogeneous ideal @xmath158 $ ] by using a variant of @xmath159/@xmath0 algorithm ( algo . [ algo : matrixf5 ] ) . 3 . dehomogenize the output to obtain a sgb of the ideal @xmath160 $ ] ( prop . [ prop : deshom ] ) ; 4 . use a sparse variant of fglm to obtain a @xmath3-dim . triangular system ( hence containing a univariate polynomial ) whose solutions are the image of the toric solutions of @xmath15 by monomial maps ( algo . [ algo : sparse - fglm ] ) ; 5 . compute the non - zero roots of the univariate polynomial and invert the monomial map to get the solutions . we focus on the four first steps of this process . the fifth step involves computing the roots of a univariate polynomial , for which dedicated techniques exist and depend on the field @xmath11 . it also involves inverting a monomial map , which can be achieved by solving a consistent linear system of @xmath161 equations in @xmath35 unknowns . in the sequel of this section , we investigate the behavior of sparse grbner bases under homogenization and dehomogeneization ( steps 1 and 3 ) . we refer the reader to ( * ? ? ? 2 ) for geometrical aspects of projective toric varieties and their affine charts . if @xmath18 verifies the assumptions of def . [ def : affinesgp ] , then there is a canonical dehomogenization map : [ def : dehom ] with the notations of def . [ def : affinesgp ] , there is a _ dehomogeneization morphism _ @xmath162 defined by @xmath163&\rightarrow&k[{s}_m]\\ & x^{(s , d)}&\mapsto&x^{s } \end{array}\ ] ] [ def : hom ] with the notations of def . [ def : affinesgp ] , for any @xmath164 $ ] , we call degree of @xmath165 , the number @xmath166_d\neq \emptyset\}$ ] . moreover the set @xmath167_{\deg(f)}$ ] contains a unique element , called the _ homogenization _ of @xmath165 . the only statement to prove is that @xmath167_{\deg(f)}$ ] contains a unique element . let @xmath168_{\deg(f)}$ ] . then @xmath169 , which implies @xmath170 . the next step is to prove that dehomogenizing a homogeneous grbner basis ( with respect to a graded ordering ) gives a grbner basis of the dehomogenized ideal . an admissible monomial ordering @xmath135 on @xmath25 $ ] is called _ graded _ if there exists an associated ordering @xmath171 on @xmath26 $ ] such that @xmath172 [ prop : deshom ] let @xmath173 be an homogeneous sgb of an homogeneous ideal @xmath174 $ ] with respect to a graded ordering . then @xmath175 is a sgb of @xmath176 with respect to the associated ordering on @xmath26 $ ] . first , notice that @xmath162 commutes with leading monomials on homogeneous components of @xmath25 $ ] : for any @xmath177_d$ ] , @xmath178 . let @xmath179 and @xmath180 be a homogeneous polynomial such that @xmath165 is equal to @xmath181 . consequently , there exists @xmath182 such that @xmath183 divides @xmath184 . applying @xmath162 , we obtain that @xmath185 divides @xmath186 . therefore @xmath175 is a sgb of @xmath176 for the associated ordering . as pointed out in @xcite , classical grbner bases algorithms are related to linear algebra via the macaulay matrices . since @xmath25 $ ] is generated by elements of degree @xmath72 , the following proposition shows that similar matrices can be constructed in the case of semigroup algebras : any monomial of degree @xmath187 in @xmath25 $ ] is equal to a product of a monomial of degree @xmath188 by a monomial of degree @xmath72 . with the notations of def . [ def : affinesgp ] , @xmath25 $ ] has the following property : for any @xmath189_d$ ] , and for all @xmath190 , there exists @xmath191_\ell$ ] s.t . this leads to the following definition of a @xmath193-grbner basis : let @xmath174 $ ] be a homogeneous ideal and @xmath135 be an admissible monomial ordering on @xmath26 $ ] . then a finite subset @xmath141 is called a @xmath193-sgb of @xmath142 if for any homogeneous polynomial @xmath194 with @xmath195 , there exists @xmath182 such that @xmath183 divides @xmath137 . note that for any @xmath196 there always exists a homogeneous @xmath193-sgb of @xmath142 . a @xmath193-sgb of @xmath142 can be deduced from a row echelon basis of the @xmath11-vector space @xmath197_d$ ] , and can be computed via the macaulay matrix : let @xmath198 be homogeneous polynomials in @xmath25 $ ] . then the _ macaulay matrix _ in degree @xmath74 of @xmath199 is a matrix with @xmath200}(d-\deg(f_i)),0)$ ] rows , @xmath201}(d)$ ] columns and entries in @xmath11 ( where @xmath202}$ ] is the hilbert function of @xmath203 $ ] ) . rows are indexed by the products @xmath204 where @xmath205 $ ] . columns are indexed by monomials of degree @xmath187 and are sorted in decreasing order w.r.t . an admissible monomial ordering . the entry at the intersection of the row @xmath204 and the column @xmath206 is the coefficient of @xmath206 in @xmath204 . by a slight abuse of notation , we identify implicitely a row in the macaulay matrix of degree @xmath187 with the corresponding polynomial in @xmath25_d$ ] . the relation between the macaulay matrix and a @xmath193-sgb is given by : [ defprop : dwit ] let @xmath207 $ ] be homogeneous polynomials , @xmath135 a graded monomial ordering , and for @xmath74 , let @xmath208 be the set of polynomials corresponding to the rows of the reduced row echelon form of the macaulay matrix in degree @xmath187 of @xmath209 . then we have @xmath210 the smallest integer @xmath38 such that @xmath211 is a sgb of the ideal @xmath212 is called _ the witness degree _ and noted @xmath213 . the first statement ( @xmath214 ) follows from the fact that @xmath208 is a triangular basis of the vector space @xmath203_d$ ] . the second statement is deduced from the inclusions @xmath215_0)\subset \chi_m(k[s_m^{(h)}]_1)\subset\dots$ ] . let @xmath173 be a sgb of @xmath216 . then @xmath213 is bounded above by @xmath217 and is therefore finite . as in the original @xmath0 algorithm @xcite , many lines are reduced to 0 during row - echelon form computations of macaulay matrices . the @xmath0 criterion @xcite ( * ? ? ? 6 ) extends without any major difficulty in this context and identifies all reductions to zero when the input system is a regular sequence in @xmath25 $ ] : [ lem : f5crit ] with the notations of algorithm [ algo : matrixf5 ] , if @xmath218 is the leading monomial of a row in @xmath219 then the polynomial @xmath220 belongs to the vector space @xmath221_{d - d_i } \ \text{and}\ u\prec m\}).\ ] ] [ algo : matrixf5 ] a direct consequence of this lemma is : algorithm [ algo : matrixf5 ] is correct . with the notations of algo . [ algo : matrixf5 ] a direct induction on @xmath187 and @xmath105 with lemma [ lem : f5crit ] shows that the row span of @xmath222 is equal to the row span of the macaulay matrix in degree @xmath187 of @xmath223 . defprop . [ defprop : dwit ] concludes the proof . in practice , the choice of the parameter @xmath193 in algorithm [ algo : matrixf5 ] is driven by the explicit bounds on the witness degree that we shall derive in section [ sec : complexity ] . the fglm algorithm and its variants might be seen as a tool to change the representation of a @xmath3-dimensional ideal . it relies on the notion of _ normal form _ relative to an ideal @xmath142 . a normal form relative to @xmath142 is a @xmath11-linear map @xmath224\rightarrow k[s]$ ] whose kernel is @xmath225 . it sends every coset of @xmath142 to the same representative , allowing effective computations in the ring @xmath226/i$ ] . one important feature of a sparse grbner basis is that it provides a normal form and an algorithm to compute it by successive reductions of leading monomials . let @xmath227 be the hilbert basis of a semigroup @xmath144 . given new indeterminates @xmath228 , any monomial in @xmath8 $ ] is the image of a monomial in @xmath229 $ ] via the morphism @xmath230\rightarrow k[{s}]$ ] defined by @xmath231 . given an admissible monomial ordering @xmath232 on the ring @xmath233 $ ] , an ideal @xmath138 $ ] and a normal form relative to @xmath142 ( given for instance by a sparse grbner basis of @xmath142 ) , algorithm [ algo : sparse - fglm ] computes a grbner basis of @xmath234 . note that @xmath235 , where @xmath236 is the map @xmath237 . also , we would like to point out that algorithm [ algo : sparse - fglm ] does not depend on the support of the input sparse system , but only on the ambient semigroup @xmath238 . the main principle of algorithm [ algo : sparse - fglm ] is similar to the original fglm algorithm @xcite : we consider the monomials in @xmath239 $ ] in increasing order until we obtain sufficiently many linear relations between their normal forms . the only difference is that the computations of the normal forms are performed in @xmath8 $ ] ( using a previously computed sparse grbner basis ) via the morphism @xmath240 . for solving sparse systems , we choose the _ lexicographical ordering _ for @xmath232 . [ algo : sparse - fglm ] @xmath241 $ ] @xmath242 $ ] @xmath243 $ ] @xmath244 $ ] return @xmath173 [ th : fglmcorrectness ] algorithm sparse - fglm is correct : it computes the reduced gb of the ideal @xmath245 $ ] with respect to @xmath232 . let @xmath246 be the output of algo . [ algo : sparse - fglm ] . set @xmath247 . first , we prove that @xmath248 . notice that each @xmath249 is of the form @xmath250 , where @xmath251 . consequently , @xmath252 and hence @xmath253 . next , let @xmath254 $ ] be a polynomial such that @xmath255 . up to reducing its nonleading monomials by @xmath173 , we can assume w.l.o.g . that all its monomials do not belong to @xmath256 . therefore , the normal forms of the images by @xmath240 of all the monomials in the support of @xmath257 are linearly independent in @xmath226/i$ ] ( otherwise the linear relation would have been detected by algo . [ algo : sparse - fglm ] ) , which means that @xmath258 and hence @xmath259 , which proves that @xmath173 is a grbner basis of @xmath234 . the proof that @xmath173 is reduced is similar . this section is devoted to the complexity of algorithms [ algo : matrixf5 ] and [ algo : sparse - fglm ] when the input is a homogeneous regular sequence . in the case of polytopal algebras , the complexity bounds of theorems [ theo : complf5 ] and [ theo : complfglm ] depend mainly on the intrinsic combinatorial properties of the defining polytope . * complexity model . * all the complexity bounds count the number of arithmetic operations @xmath260 in @xmath11 ; each of them is counted with unit cost . it is not our goal to take into account operations in the semigroup @xmath61 . the first goal is to bound @xmath213 ( see defprop . [ defprop : dwit ] ) via the hilbert series of @xmath8/i$ ] . for regular sequences , this hilbert series can be computed by the following classical formula : [ prop : hssuitereg ] let @xmath2 be a normal lattice polytope , @xmath261 $ ] be a homogeneous regular sequence of homogeneous polynomials of respective degrees @xmath262 and @xmath263 $ ] . then @xmath115/i}(t)=\operatorname{hs}_{{\mathscr p}}(t)\cdot\prod_{i=1}^p ( 1-t^{d_i}).\ ] ] see _ e.g. _ ( * ? ? ? * exercise 21.17b ) . the next lemma gives an explicit bound for the witness degree of regular sequences in a normal polytopal algebra : [ lem : sparsedreg ] let @xmath1 be a normal lattice polytope and @xmath264 be a homogeneous regular sequence in @xmath92 $ ] of degrees @xmath265 . then any @xmath266)+1+\sum_{j=1}^n(d_j-1)\right]$]-sgb of the ideal @xmath267 is a sgb of @xmath142 . in other words @xmath268)+1+\sum_{j=1}^n(d_j-1)$ ] . by prop . [ prop : hssuitereg ] and with the notations of prop . [ prop : hspolytope ] , the hilbert series of @xmath92/i$ ] is equal to @xmath269 where @xmath270 $ ] is a univariate polynomial with @xmath271)-1+\sum_{i=1}^p(d_i-1)$ ] . now , notice that the hilbert series of @xmath92/i$ ] is equal to that of @xmath92/\operatorname{lm}(i)$ ] . therefore @xmath272/\operatorname{lm}(i)}(d)$ ] is constant for @xmath273 . since @xmath274 implies @xmath275 , we obtain @xmath276{@{}l@ { } } s\in ( d\cdot{\mathscr p})\cap{\mathbb{z}}^n \text { and}\\ s\notin ( ( d-1)\cdot{\mathscr p})\cap{\mathbb{z}}^n \ } \end{array}\\=\deg(k(t))+1 . \end{array}\ ] ] consequently , minimal generators of @xmath139 and hence minimal homogeneous grbner bases of @xmath142 have degree at most @xmath277)+1+\sum_{j=1}^n(d_j-1)$ ] . now that we have an upper bound for the witness degree , we can estimate the cost of computing a sgb by reducing the macaulay matrix in degree @xmath213 ( although sparse - matrixf5 is a much faster way to compute a sgb in practice , it is not easy to bound precisely its complexity ) . note that @xmath121)$ ] in the following theorem can be deduced from prop . [ prop : regpolytope ] . [ theo : complf5 ] with the same notations as in lemma [ lem : sparsedreg ] , the complexity of computing a sgb of @xmath278 $ ] by reducing the macaulay matrix in degree @xmath213 is bounded above by @xmath279 where @xmath54)+1+\sum_{j=1}^n(d_j-1)$ ] and @xmath280 is a feasible exponent for the matrix multiplication ( @xmath53 with @xcite ) . let @xmath281 $ ] be the ideal generated by @xmath282 . the number of columns and rows of the macaulay matrix in degree @xmath187 are respectively @xmath283 consequently , the row echelon form of such a matrix can be computed within @xmath284 field operations ( * ? ? ? 2.11 ) . by proposition [ prop : deshom ] and lemma [ lem : sparsedreg ] , for @xmath285)+1+\sum_{j=1}^n(d_j-1)$ ] , this provides a sgb of @xmath286 . we now investigate the complexity of algorithm [ algo : sparse - fglm ] when @xmath138 $ ] is a @xmath3-dim . ideal , and use the same notations as in section [ subsec : sparse - fglm ] . notice that the map @xmath240 induces an isomorphism @xmath287/\varphi^{-1}(i ) \rightarrow k[{s}]/i$ ] and therefore algorithm [ algo : sparse - fglm ] may be seen as a way to change the representation of @xmath8/i$ ] . [ theo : complfglm ] set @xmath288/i)$ ] and let @xmath289 be the cardinality of the hilbert basis of @xmath61 . if the input normal form is computed via a reduced sgb of @xmath138 $ ] ( for some monomial ordering ) , @xmath61 is a simplicial affine semigroup ( see def . [ def : simplicial ] ) and @xmath8 $ ] is cohen - macaulay , then algorithm [ algo : sparse - fglm ] computes the grbner basis @xmath173 with at most @xmath290 operations in @xmath11 . once the @xmath289 matrices of size @xmath291 representing the multiplications by @xmath292 in the canonical monomial basis of @xmath226/i$ ] are known , step * ( 1 ) * in algorithm [ algo : sparse - fglm ] can be achieved in @xmath293 as in the classical fglm algorithm @xcite . steps * ( 2 ) * and * ( 3 ) * are done by linear algebra as in @xcite , which leads to a total complexity of @xmath290 since the same analysis holds . it remains to prove that the multiplication matrices can be constructed in @xmath290 operations ( this is a consequence of ( * ? ? ? 2.1 ) in the classical case ) . since @xmath8 $ ] is cohen - macaulay and @xmath61 is simplicial , we obtain by ( * ? ? ? * thm . 1.1 ) that for any two distinct @xmath294 and for any @xmath13 , if @xmath295 and @xmath296 are in @xmath61 then @xmath297 . with this extra property , the proof of ( 2.1 ) extends to semigroup algebras . if the input system is a regular sequence of laurent polynomials , then @xmath298 can be bounded by the mixed volume of their newton polytopes by kushnirenko - bernstein s theorem @xcite . in this section , we specialize theorems [ theo : complf5 ] and [ theo : complfglm ] to several semigroups to obtain new results on the complexity of solving inhomogeneous systems with classical gb algorithms ( @xmath2 is the standard simplex ) , multi - homogeneous systems ( @xmath2 is a product of simplices ) and we state a variant of frberg s conjecture for overdetermined sparse systems . * inhomogeneous dense systems . * if @xmath299 is the standard simplex in @xmath83 , then computations of a sparse grbner basis in the cone over @xmath300 correspond to classical grbner bases computations using the so - called `` sugar strategy '' introduced in @xcite . the following corollary shows that specializing theorems [ theo : complf5 ] and [ theo : complfglm ] with @xmath299 recovers the usual complexity estimates for grbner bases computations let @xmath264 be a regular sequence of polynomials of respective degrees @xmath265 in @xmath301 $ ] . then the complexity of computing a classical grbner basis of @xmath302 with respect to a graded monomial ordering is bounded by @xmath303 where @xmath304 . in particular , notice that if @xmath264 are inhomogeneous polynomials in @xmath81 $ ] , then their homogeneous counterparts in @xmath301 $ ] form a regular sequence if and only if their `` classical homogenization '' form a regular sequence . * multi - homogeneous systems . * another class of polynomials appearing frequently in applications are _ multi - homogeneous systems_. a polynomial of multi - degree @xmath305 w.r.t . a partition of the variables in blocks of sizes @xmath306 is a polynomial whose newton polytope is included in @xmath307 . in that case , the associated polytope is a product of simplices , which allows us to state the following complexity theorem : [ theo : multihom ] let @xmath264 be a regular sequence of polynomials of multi - degree @xmath305 w.r.t . a partition of the variables in blocks of sizes @xmath306 ( with @xmath308 ) . then the combined complexity of steps ( 1 ) to ( 4 ) of the solving process in section [ sec : sgb ] is bounded by @xmath309 applying theorems [ theo : complf5 ] and [ theo : complfglm ] with @xmath2 equal to @xmath310 yields the complexity bound in terms of @xmath213 , @xmath311 and @xmath298 . first , notice that the semigroup generated by @xmath34 is @xmath151 , and hence @xmath312 . next , @xmath313 has an interior lattice point if and only if for all @xmath105 , @xmath314 has an interior lattice point , _ i.e. _ @xmath315 . the smallest @xmath316 that verifies this condition is @xmath317 . by prop . [ prop : regpolytope ] , @xmath121)=n+1-\max(\lceil ( n_1 + 1)/d_1\rceil,\ldots,\lceil ( n_\ell+1)/d_\ell\rceil)$ ] . since the @xmath264 have degree @xmath72 in @xmath92 $ ] , we get @xmath268)+1 $ ] . finally , notice that the unnormalized volume of @xmath318 is @xmath319 . consequently , the unnormalized volume of @xmath2 is @xmath320 . normalizing the volume amounts to multiplying this value by @xmath88 , which yields the formula for @xmath321 and equals the multi - homogeneous bzout number . the number of solutions ( counted with multiplicity ) is classically bounded by this value and hence @xmath322 . finally , we state a variant of frberg s conjecture @xcite in the sparse framework , leading to a notion of `` sparse semi - regularity '' . it provides a bound on the witness degree of generic overdetermined sparse systems : this conjecture can be used to adjust the parameter @xmath193 of algorithm [ algo : matrixf5 ] . [ conj : froberg ] let @xmath1 be a normal lattice polytope , @xmath323 be a sequence of integers with @xmath324 . if @xmath325 $ ] are generic homogeneous polynomials of respective degrees @xmath326 , then @xmath327/\langle f_1,\ldots , f_m\rangle}(t)=\left[\operatorname{hs}_{{\mathscr p}}(t)\prod_{i=1}^m ( 1-t^{d_i})\right]_+,\ ] ] where @xmath328_+$ ] means truncating the series expansion at its first nonpositive coefficient . systems for which this equality holds are called _ semi - regular_. the witness degree of a semi - regular sequence is bounded above by the index of the first zero coefficient in the series expansion of @xmath329/\langle f_1,\ldots , f_m\rangle}(t)$ ] . in this section , we estimate the speed - up that one can expect for solving sparse systems or systems of laurent polynomials via sparse grbner bases computations , compared to classical grbner bases algorithms . the same linear algebra routines are used in the compared implementations . consequently , the speed - up reflects the differences between the characteristics ( size , sparseness , ) of the matrices that have to be reduced . we compare sparse - matrixf5 ( abbreviated sp - matrixf5 ) with the implementation of the @xmath0 algorithm in the fgb library . we report more detailed experimental results on a benchmarks webpage . in all these experiments , the base field @xmath11 is the finite field @xmath330 . all tests are done with overdetermined systems with one rational solution in @xmath331 . the goal is to recover this solution . in that case , the fglm algorithm is not necessary since the sparse grbner basis describes explicitly the image of the solution by a monomial map . in several settings , we report the speed - up obtained with our prototype implementation . * bilinear systems . * in table [ table : bilinear ] , we focus on overdetermined bilinear systems . for @xmath332 , we generate a system of @xmath218 polynomials with support @xmath333 uniformly at random in the set of such systems which have at least one solution in @xmath334 . * systems of bidegree @xmath335 . * in table [ table : bidegree ] , we report the performances on overdetermined systems with support @xmath336 . note that we obtain important speed - ups when @xmath337 ( more than 19000 for @xmath338 ) . * fewnomial systems . * in table [ table : fewnomials ] , we report performances on fewnomial systems . the complexity analysis in section [ sec : complexity ] do not apply to this context because the semigroup algebra in which we compute is not normal . however , the correctness of the algorithms still holds . the systems are generated as follows : for @xmath339 we pick @xmath340 monomials of degree @xmath341 in @xmath35 variables uniformly at random and we generate a system of @xmath218 polynomials with this support in @xmath342 $ ] with random coefficients such that there is at least one solution in @xmath331 . the computations are done w.r.t . the semigroup generated by the @xmath340 monomials . note that for some specific instances , the speed - up factor can be as high as 16800 .	toric ( or sparse ) elimination theory is a framework developped during the last decades to exploit monomial structures in systems of laurent polynomials . roughly speaking , this amounts to computing in a _ semigroup algebra _ , _ i.e. _ an algebra generated by a subset of laurent monomials . in order to solve symbolically sparse systems , we introduce _ sparse grbner bases _ , an analog of classical grbner bases for semigroup algebras , and we propose sparse variants of the @xmath0 and fglm algorithms to compute them . our prototype `` proof - of - concept '' implementation shows large speed - ups ( more than 100 for some examples ) compared to optimized ( classical ) grbner bases software . moreover , in the case where the generating subset of monomials corresponds to the points with integer coordinates in a normal lattice polytope @xmath1 and under regularity assumptions , we prove complexity bounds which depend on the combinatorial properties of @xmath2 . these bounds yield new estimates on the complexity of solving @xmath3-dim systems where all polynomials share the same newton polytope ( _ unmixed case _ ) . for instance , we generalize the bound @xmath4 on the maximal degree in a grbner basis of a @xmath3-dim . bilinear system with blocks of variables of sizes @xmath5 to the multilinear case : @xmath6 . we also propose a variant of frberg s conjecture which allows us to estimate the complexity of solving overdetermined sparse systems .
in the context of classical field theories , topological structures , such as kinks @xcite , vortices @xcite and magnetic monopoles @xcite , are described as static solutions to some nonlinear models . in particular , such models must allow for the spontaneous symmetry breaking mechanism , since topological solutions are formed during a symmetry breaking phase transition . in this sense , topologically non - trivial configurations are of great interesting to physics @xcite , mainly concerning the cosmological consequences they may engender , since such configurations can appear in a rather natural way , during phase transitions in the early universe . in particular , vortices are described as rotationally symmetric solutions of a planar abelian - maxwell - higgs model endowed by a fourth - order scalar potential for the matter self - interaction , which also introduces symmetry breaking and nonlinearity . in the usual case , the maxwell term controls the dynamics of the gauge field , and the covariant derivative squared term controls the dynamics of the scalar field . in such context , vortices are finite - energy solutions of a set of two coupled first - order differential equations , named bogomolnyi - prasad - sommerfield ( bps ) equations @xcite . in this case , the bps vortices are the minimal - energy solutions of the model , and they have interesting applications , mainly concerning the superconductivity phenomena @xcite and the superfluid he@xmath3 . during the last years , beyond the standard configurations previously cited , modified ones , also named _ topological k - solutions _ , have been intensively studied . these solutions arise in a special class of theoretical field models , generically named _ k - theories _ , which are endowed by non - usual kinetic terms . as expected , such terms change the dynamics of the overall model under investigation . moreover , it is important to reinforce that the idea of a non - standard dynamics arises in a rather natural way , in the context of string theories . in general , k - theories have been used as effective models mainly in the cosmological scenario . here , many authors have used the so - called k - essence models @xcite to investigate the present accelerated inflationary phase of the universe @xcite . also , such models can be used to study strong gravitational waves @xcite , dark matter @xcite , tachyon matter n10 and others @xcite . there are also important motivations concerning the strong interaction physics ; see , for instance , ref . @xcite . in this context , the overall conclusion is that the existence of topological structures is quite sensible to the use of non - standard kinetic terms . here , the more interesting issue is that such structures can exist even as solutions of some theories which do not allow for the spontaneous symmetry breaking mechanism @xcite . on the other hand , a rather natural way to study such topological solutions is comparing them with their canonical counterparts . in this sense , some of us have already investigated topological solutions in the context of k - field models endowed by spontaneous symmetry breaking potentials for the scalar - matter self - interaction , and interesting results can be found , for instance , in ref . other interesting results concerning k - field models and their classical solutions can be found , for instance , in ref . @xcite . another interesting issue concerning topological k - solutions is that they can be either much larger or much smaller than their usual counterparts . in this sense , k - bosons can mediate either large - range or small - range interactions . also , important physical quantities , such as energy density and electric and magnetic fields , can exhibit proeminent variations on their tipical profiles , including on their maximum values ; for a detailed treatment of some of such features , see ref . @xcite . in the present context , some k - theories can also support topological solutions with a finite wavelength . these solutions , generically named _ @xcite , are quite different from the standard topological structures , which interact even if separated by an infinite displacement , since they have an infinite wavelength : two adjacent compactons will interact only if they come into close contact , due to their already explained finite wavelength . in this sense , compactons are most appropriated to describe particle - like configurations than the usual non - compactified topological structures . so , in this work we present new results concerning the topological solutions of an abelian - maxwell - higgs model endowed by non - usual kinetic terms to both gauge and scalar fields . here , the non - standard dynamics are introduced by two positive functions , @xmath4 and @xmath5 , which couple with the maxwell term and with the covariant derivative squared term , respectively . the euler - lagrange equations of motion of such theory are hard to solve , then we focus our attention only on the finite - energy solutions of the bps equations of the model . these equations can be obtained by minimizing the energy functional of the model , which can be done via an important constraint between @xmath4 , @xmath6 and the scalar potential @xmath7 for the matter self - interaction ; see eqs . and below . in this case , rotationally symmetric bps vortices are described as the minimal - energy solutions of the cited model , and they engender interesting features , as explained below . this paper is outlined as follows : in the next sec.[general ] , we introduce the model and develop the theoretical framework which allows us to get to its bogomolnyi - prasad - sommerfield equations . in sec.[numerical ] , we prove the consistence of the theoretical framework previously developed by using it to investigate the existence of new bps states . here , we note that such states are constrained to the choices made for @xmath8 and @xmath5 : for any acceptable pair of such functions , there is a corresponding bpsconfiguration . also in sec.[numerical ] , we show how to use the theoretical framework presented in this work to recover the recover the bpsresults concerning the standard maxwell - higgs model . in section numerical2 , we perform the numerical analysis concerning the new bpsstates previously presented , we depict the corresponding minimal - energy modified solutions and comment on their main features . finally , in sec.end , we present our conclusions and perspectives . from now on , we use standard conventions , including natural units system , and a plus - minus signature for the planar minkowski metric : diag@xmath9 . in this section , we introduce the model . it is described by the @xmath10-dimensional lagrange density@xmath11here , @xmath12 is the usual faraday field strength tensor , @xmath13 is the covariant derivative and @xmath7 is the spontaneous symmetry breaking potential . also , @xmath4 is the `` dieletric function '' , and @xmath14 stands for the non - standard kinetic matter term . both @xmath4 and @xmath1 are dimensionless functions to be given below , as functions of the amplitude of the scalar field . in the case of vortices , it is convenient to deal with dimensionless variables . so , for simplicity , we introduce the mass scale @xmath15 , and use it to implement the scale transformations : @xmath16 , @xmath17 , @xmath18 , @xmath19 and @xmath20 , where @xmath21 stands for the symmetry breaking parameter of the model . in this case , we get @xmath22 , with @xmath23 being the dimensionless lagrange density to be used from now on , which has the same functional form of @xmath24 . moreover , we omit the coupling constant related to the scalar matter self - interaction . also , we take @xmath25 and @xmath21 as real and positive parameters . to search for vortex solutions , the canonical procedure is to deal with the euler - lagrange equations of the model . in the present case , these equations are@xmath26and@xmath27here , we take @xmath28 and@xmath29as the modified 4-current vector . the gauss law for time - independent fields can be written as@xmath30where @xmath31 . we note that the eq . is trivially verified by @xmath32 . so , we fix this gauge and use it from now on . we look for vortex solutions of the form@xmath33@xmath34here , @xmath35 and @xmath36 are polar coordinates , and @xmath37 is the winding number ( vorticity ) of the solution . in terms of and , the euler - lagrange equations and can be rewritten as@xmath38@xmath39equations and are the euler - lagrange equations of motion to the profile functions @xmath40 and @xmath41 , respectively . to solve and , we need to specify the model . in general , we can do it by choosing non - trivial forms to @xmath42 and @xmath43 . in this case , we have to keep in mind that both these functions must be positive , in order to avoid problems with the energy of the model ; see the expression for the energy density below . also , we need to choose a higgs potential @xmath44 which allows for the spontaneous symmetry breaking mechanism . before that , we note that the limit @xmath45 leads us back to the model studied in @xcite , which is supported by applications concerning the interaction between quarks and gluons @xcite . in this case , the limit @xmath46 leads us back to the usual maxwell - higgs theory . in this context , if we choose the symmetry breaking higgs potential@xmath47the euler - lagrange equations and can then be rewritten as@xmath48@xmath49according to our conventions , eqs . and are completely solvable by the first order differential equations@xmath50@xmath51 the solutions of eqs . and are the bogomolnyi - prasad - sommerfield ( bps ) states of the standard maxwell - higgs model . these solutions are the well - known abrikosov - nielsen - olesen ( ano ) ones @xcite , which solve the equations of motion and by minimazing the energy of the resulting vortex configurations . in general , for non - trivial choices to @xmath42 and @xmath52 , the equations of motion and will be much more sofisticated than the eqs . and . so , in order to get an useful insight about the non - trivial case , we consider the expression for the energy density of the modified vortex solutions:@xmath53from eq . , we note that the presence of non - trivial @xmath54 and @xmath43 makes it hard to obtain the bps states of the modified model . even in this case , the existence of such states is still possible , and it is closely related to an important constraint between @xmath42 , @xmath43 and @xmath44:@xmath55as we clarify below , for any positive choices to @xmath42 and @xmath43 , there is a corresponding symmetry breaking higgs potential @xmath44 as a solution of . in this context , generalized first order equations can be found by minimizing the energy functional . to search for bps vortex solutions in the modified model , we need to know how the functions @xmath41 and @xmath40 behave , near the origin and asymptotically . near the origin , these functions must avoid singular fields . so , they have to behave according to@xmath56where we have fixed @xmath57 , for simplicity . also , the symmetry breaking vortex configurations must have a finite total energy . so , as a condition to make the energy finite , the energy density must vanish . then , asymptotically , @xmath41 and @xmath40 have to obey@xmath58 in the next sec . iii , we use the theoretical framework developed in this section to investigate the existence of bps states in the generalized model . also , we present and comment the resulting numerical solutions . finally , we show how to map the standard maxwell - higgs and chern - simons - higgs first order equations ; even in these cases , eq . leads us to a physically different rotationally symmetric solutions . we now pay due attention to the modified bps states themselves . here , we choose positive functional forms to @xmath42 and @xmath52 . then , we solve the constraint to get to the consistent higgs potential @xmath44 , which must allow for the spontaneous symmetry breaking mechanism . a posteriori , we use these conventions to obtain the first order differential equations of the modified model , by minimizing its energy functional . then , in the next sec.iv , we numerically solve these equations , according to the boundary conditions and . finally , in sec.v , we comment the main features of the resulting solutions . the model to be studied here has two unusual functions , @xmath42 and @xmath43 , which we have to fix to determine how they affect the vortex solutions . the function @xmath42 stands for a `` dieletric function '' , and it controls the non - standard kinetic term to the gauge field . on the other hand , the function @xmath43 controls the non - usual kinetic scalar matter term . both these functions are dimensionless , and are functions of the amplitude of the scalar field . also , they must be positive , in order to avoid problems with the energy of the model ; see eq . . a rather natural way to investigate the consistence of the theoretical framework developed in the previous sec . ii is to study the bps states of the standard maxwell - higgs model , which is defined by @xmath59 . in this case , eq . leads us to the potential . then , the energy density can be rewritten as@xmath60the resulting total energy @xmath61 is minimized by the first order equations and . in this context , the energy of the bps states is@xmath62as usual , this is the lower bound of the energy functional , i.e. , the bogomolnyi bound . also , it is quantized according to the winding number @xmath63 . now , we introduce an interesting new model , which is defined by@xmath64according to these choices , we solve to get to the potential@xmath65which allows for the spontaneous symmetry breaking mechanism , as desired . we point out that this new model is not a parametrization of the standard maxwell - higgs one , since the vacuum manifold of the two models are quite different : it is a circle for the usual model , while it is a dot surrounded by a circle for the new model . using eqs . and , the energy density can then be rewritten as@xmath66then , the corresponding energy functional is minimized by the equations@xmath67@xmath68we note that the modified equations and can not be parametrized into the standard ones , i.e. , eqs . and . even in this case , the energy of the modified field solutions is bounded from below , and the bogomolnyi bound is@xmath69here , as in the standard case , the total energy of the bps states is quantized ; see eq . . in order to reinfoce the consistence of the theoretical framework presented in this work , we introduce another modified model . it is defined by@xmath70according to these conventions , eq . leads us to the standard higgs potential . so , the modified model has the same vacuum structure as the usual model . even in this context , the new model is not a parametrization of the standard one ; see the bps eqs . and below . the modified energy density can be written in the form@xmath71and the resulting energy functional is minimized by the first order equations@xmath72@xmath73then , the bogolmolnyi bound is also given by , and we conclude that the modified solutions and have the same energy of the standard and ones . to end this section , we show how to use the theoretical framework developed in this work to map the standard maxwell- and chern - simons - higgs first order differential equations . we do it by reviewing the model studied in @xcite , which is defined by @xmath74 . in this case , as presented in that work , an interesting choice to the dieletric function is@xmath75here , @xmath76 is a real auxiliary parameter which controls the model , and @xmath77 stands for the coupling constant corresponding to the chern - simons term . in the limit @xmath78 , eq . gives @xmath79 , and reproduces the standard maxwell - higgs system . on the other hand , the limit @xmath80 gives@xmath81we then solve eq . to get@xmath82which is exactly the sixth - order symmetry breaking higgs potential one finds in the usual self - dual chern - simons - higgs model . now , from eqs . and , the energy density can be written in the form@xmath83the resulting total energy is minimized by the differential equations@xmath84@xmath85and the bogomolnyi bound is @xmath86 . the equations and are just the bps ones for the standard chern - simons - higgs system . in this sense , eq . works as a unified way to map both the maxwell- and the chern - simons - higgs first order differential equations . finally , we point out that , despite the possibility of reproducing such chern - simons equations , the unified picture leads us to a physically different situation , since the modified model we consider here only supports non - charged field solutions ; see the gauss law above . in this case , the modified nontopological field solutions are unstable , and they can decay into the elementary mesons of the model . this fact makes an important difference , since the nontopological configurations presented by the self - dual chern - simons - higgs theory are stable , and can not decay . in the next section , we present the modified first order numerical solutions for the profile funtions @xmath41 and @xmath87 . also , we comment on the main features of such solutions . let us focus attention on the modified numerical solutions themselves . the equations to be solved are the first order ones and , and and . also , we solve the standard equations and , for comparison . in all cases , the functions @xmath41 and @xmath40 must obey the boundary conditions and . in the present work , the numerical strategy to be employed is the relaxation one . in this case , it is necessary to input an approximated field solution . then , our algorithm will `` relax '' it into the correct one . to start the numerical analysis , we consider a variation of the standard maxwell - higgs model , in which the potential for the self - interaction for the scalar field is not present . then , we use its solutions to solve the self - dual maxwell - higgs case , i.e. , the equations and , from which we get the well - understood abrikosov - nielsen - olesen ( ano ) configurations . we then use these solutions to initialize the numerical study of the modified theory . starting from such ano configurations , we numerically solve the modified bps equations presented in the previous section , i.e. , eqs . and , and also eqs . and , for @xmath88 the solutions for the profile functions @xmath41 and @xmath40 are plotted in figs . 1 and 2 , respectively . also , we depict the solutions for the corresponding energy densities ; see figure 3 . for the models ( dash - dotted red line ) and ( dashed blue line ) . the standard solution is also depicted ( solid black line ) , for comparison.,width=321 ] in figure 1 , we present the solutions for the profile function @xmath89 , and we see that both the modified profiles reach their vacuum values more slowly than their canonical counterpart . in this sense , such solutions have a core which is greater than the usual maxwell - higgs one . here , the conclusion is that , in general , the introduction of a non - canonical dynamics allows for the existence of self - dual field solutions @xmath41 with a increased characteristic length . also , we note that the solution for and goes to its vacuum configuration more slowly than that for and . so , beyond the fact of increasing the core of the solutions , we note that increases it more than the . we believe that this fact is related with the corresponding self - dual higgs potential , which is of sixth - order for the model , and of fourth - order for the one ; see equations and above . . conventions as in fig . 1.,width=321 ] in fig . 2 , we depict the numerical solutions for the function @xmath90 . here , as in the fig . 1 , we note that the modified solutions go to their vacuum states more slowly than the standard maxwell - higgs profile , and so such solutions have a increased characteristic length . this behaviour reinforces the previous conclusion , which states that a non - standard dynamics leads to a self - dual field solutions with a increased core . we also see that , beyond the fact of increasing the core of @xmath40 , the model enlarges it more than the . finally , we point out the existence of a proeminent plateau in the profile related to , near the origin : such structure is also present in the self - dual chern - simons - higgs case , which is governed by the potential . in this context , it is interesting to note that the existence of such proeminent plateau seems to be closely related to the vacuum manifold of a sixth - order symmetry breaking potential , since it is also present in the modified model , which is governed by . the fig . 3 encloses the solutions for the energy densities of the new bps states ; see equations and . we see that the profile for the energy density corresponding to the modified model is quite different from that of the standard maxwell - higgs theory : in the canonical case , the energy density reaches its maximum value in the limit @xmath91 , and it is monotonically decrescent for all @xmath35 . on the other hand , in the modified case , the energy density reaches its maximum value at some finite distance @xmath92 from the origin , and it is not monotonically decrescent for all values of the independent variable . in this context , this difference reinforces our previous conclusion , since such modified behaviour mimics the one of the energy density of the usual self - dual chern - simons theory . . conventions as in fig . 1.,width=321 ] the solution for the bps energy density of the modified model is also depicted in figure 3 , and we see that its behaviour is qualitatively similar to that of the maxwell - higgs model : it reaches its maximum value as @xmath35 goes to @xmath93 , and it is monotonically decrescent for all @xmath35 . here , however , there are two important differences : the first one is on the maximum value of the modified profile itself , which is smaller than the usual one , and the second one is on the characteristic length of the non - standard solution , which is greater than its canonical counterpart . an important issue concerning the study of topological structures is the computation of their topological charges , which must be conserved . in the present case , i.e. , for electrically non - charged field solutions of the form and , this charge is related to the magnetic field generated by such solutions themselves . to investigate this issue , we introduce the topological current@xmath94which is clearly conserved . here , @xmath95 . the 0@xmath96-component of such current , that is , the topological charge density , can be written as@xmath97and we see that it is directly proportional to the magnetic field @xmath98 . in this case , the topological charge is given by@xmath99here , @xmath100 stands for the flux of the magnetic field . according , and , the magnetic flux @xmath101 can be rewritten in the form@xmath102and we note that the topological charge @xmath103 is conserved , and it is quantized according to the winding number @xmath63 . in this sense , to better specify the field configurations studied in this work , a rather natural way is to consider their corresponding magnetic fields . in the present context , this field is given by@xmath104 the modified numerical profiles for the magnetic field are plotted in figure 4 . also for such field , the non - usual dynamics introduced earlier leads to bps solutions with a increased core , since the modified profiles have a characteristic length which is greater than the standard one . in this case , as for the energy densities previously depicted , the magnetic field associated to is quite different from the canonical self - dual one , which reaches its maximum value for @xmath91 , and is monotonically decrescent ever : the non - standard solution reaches its maximum value at some finite distance from the origin , and it is not monotonically decrescent for all @xmath35 . such behaviour is similar to that of the self - dual magnetic field related to the canonical chern - simons - higgs model . . conventions as in fig . 1.,width=321 ] to end this section , we discuss the solution for the magnetic field related to the model . this solution is also depicted in fig . 4 , from which we note that the behaviour of such magnetic field is quite similar to that of the usual maxwell - higgs case , since it reaches its maximum value as @xmath91 , and it is monotonically decrescent ever . here , we point out the behaviour of the modified solution , which assures the conservation of the topological charge ; see eqs . and . in the present letter , we have considered the existence of rotationally symmetric bps solutions in a @xmath105-dimensional space - time . we have investigated such solutions in a modified self - dual maxwell - higgs model endowed by a non - standard dynamics . here , the modification was introduced in terms of two non - trivial functions , @xmath106 and @xmath5 , which must be positive , in order to avoid problems with the energy density of the model . so , while @xmath4 couples with the maxwell term and , as a consequence , changes the dynamics of the gauge field in a non - usual way , @xmath107 couples with the squared covariant derivative of the non - charged scalar field , then leading to a non - standard dynamics to such field . in this context , consistent first order equations were obtained since the non - trivial functions @xmath4 and @xmath5 are constrained to the symmetry breaking higgs potential @xmath108 of the modified model ; see eq . . we have integrated the bps equations by means of the relaxation method , and the numerical results we found are depicted in figs . 1 and 2 , for some interesting choices to the functions @xmath106 and @xmath5 ; see eqs . and . according these solutions , we conclude that both the profile functions @xmath41 and @xmath40 are quite sensible to the choices made for @xmath106 and @xmath5 . in particular , it is important to reinforce that each of the models and is related to a very specific symmetry breaking potential @xmath2 and , as a consequence , such models present distinct vacuum manifolds ; for details , see eqs . and . also , using the previous results we found for @xmath41 and @xmath40 , we have integrated both the energy density @xmath109 and the magnetic field @xmath110 for the modified models and , and these solutions are depicted in figs . 3 and 4 , respectively ; see eqs . , and . the numerical analysis reveals that the solutions corresponding the model behave as those predicted by the chern - simons - higgs theory , which reach their maximum values at some finite distance from the origin , and are not monotonically decrescent for all @xmath35 . on the other hand , the solutions corresponding the model behave as the maxwell - higgs ones , since they reach their maximum values as @xmath91 , and are monotonically decrescent for all values of the independent variable . so , as a consequence , we conclude that also the energy density @xmath111 and the magnetic field @xmath110 are both sensible to the choices made for @xmath106 and @xmath5 . we hope that this work may stimulate subsequent analysis in the field , concerning mainly the features that the modified solutions engender . also , we point out that the variation on the characteristic length of the modified solutions presented here is closely related to the effects of anisotropy , which is a feature tipically present in the effective field models used to describe the behaviour engendered by low - energy condensed matter systems . in this sense , we point out that such effective models are usually based on the lorentz - breaking ideia , since it introduces the issue of anisotropy explicity . in the context of lorentz - violating models , such effects where already studied by some of us ; see , for instance , ref . @xcite . a very interesting issue concerns the use of modified models to mimic the very same solutions engendered by the standard maxwell - higgs theory . in particular , such issue is now under consideration , and we hope to report new results in a near future . the authors would like to thank capes , cnpq ( brazil ) and fct project cern / fp/116358/2010 ( portugal ) for partial financial support . also , we are grateful to c. adam , f. correia and d. rubiera - garcia for useful discussions . e. da hora thanks the department of mathematical sciences of durham university ( united kingdom ) , for all their hospitality while doing this work . a. vilenkin and e. p. s. shellard , _ cosmic strings and other topological defects _ ( cambridge university press , cambridge , england , 1994 ) . n. manton and p. sutcliffe , _ topological solitons _ ( cambridge university press , cambridge , england , 2004 ) . t. h. r. skyrme , proc . a * 260 * , 127 ( 1961 ) . l. faddeev and a. j. niemi , phys . lett . * 82 * , 1624 ( 1999 ) . d. a. nicole , j. phys . g * 4 * , 1363 ( 1978 ) . s. deser , m. j. du and c. j. isham , nucl . b * 114 * , 29 ( 1976 ) . h. aratyn , l. a. ferreira and a. h. zimerman , phys lett . * 83 * , 1723 ( 1999 ) . d. bazeia , j. menezes and r. menezes , phys . lett . * 91 * , 241601 ( 2003 ) . c. adam , j. sanchez - guillen , r. a. vazquez and j. a. wereszczynski , j. math . phys . * 47 * , 052302 ( 2006 ) . d. bazeia , e. da hora , c. dos santos and r. menezes , phys . d * 81 * , 125014 ( 2010 ) . d. bazeia , e. da hora , r. menezes , h. p. de oliveira and c. dos santos , phys . rev . d * 81 * , 125016 ( 2010 ) . c. dos santos and e. da hora , eur . j. c * 70 * , 1145 ( 2010 ) ; eur . j. c * 71 * , 1519 ( 2011 ) . c. dos santos , phys . d * 82 * , 125009 ( 2010 ) . c. adam , n. grandi , j. sanchez - guillen and a. wereszczynski , j. phys . a * 41 * , 212004 ( 2008 ) . erratum - ibid . a * 42 * , 159801 ( 2009 ) . c. adam , j. sanchez - guillen and a. wereszczynski , j. phys . a * 40 * , 13625 ( 2007 ) . erratum - ibid . a * 42 * , 089801 ( 2009 ) . c. adam , n. grandi , p. klimas , j. sanchez - guillen and a. wereszczynski , j. phys . a * 41 * , 375401 ( 2008 ) . c. adam , p. klimas , j. sanchez - guillen and a. wereszczynski , j. phys . a * 42 * , 135401 ( 2009 ) . e. babichev , phys . d * 74 * , 085004 ( 2006 ) . d. bazeia , l. losano , r. menezes and j. c. r. e. oliveira , eur . j. c * 51 * , 953 ( 2007 ) . x. jin , x. li . and d. liu , classical quantum gravity * 24 * , 2773 ( 2007 ) . m. n. barreto , d. bazeia and r. menezes , phys . d * 73 * , 065015 ( 2006 ) . d. bazeia , m. m. ferreira jr . , a. r. gomes and r. menezes , physica d * 239 * , 942 ( 2010 ) . r. casana , m. m. ferreira jr . , e. da hora and c. m. cantanhede , in preparation .	we look for topological bps solutions of an abelian - maxwell - higgs theory endowed by non - standard kinetic terms to both gauge and scalar fields . here , the non - usual dynamics are controlled by two positive functions , @xmath0 and @xmath1 , which are related to the self - dual scalar potential @xmath2 of the model by a fundamental constraint . the numerical results we found present interesting new features , and contribute to the development of the recent issue concerning the study of generalized models and their applications .
in this paper , we utilize techniques inspired by riemannian geometry and the theory of stochastic processes in order to control eigenvalues of graphs . in particular , we shall quantify the deviation of a ( connected , undirected , weighted , finite ) graph @xmath1 from being bipartite ( a bipartite graph is one without cycles of odd lengths ; equivalently , its vertex set can be split into two classes such that edges can be present only between vertices from different classes ) in terms of a spectral gap . the operator whose spectrum we shall consider here is the normalized graph laplacian @xmath0 ( it is unitarily equivalent to the one studied in chung @xcite ) . this is the operator underlying random walks on graphs , and so , this leads to a natural connection with the theory of stochastic processes . we observe that on a bipartite graph , a random walker , starting at a vertex @xmath14 at time 0 and at each step hopping to one of the neighbors of the vertex where it currently sits , can revisit @xmath14 only at even times . this connection then will be explored via the eigenvalues of @xmath0 . more precisely , the largest eigenvalue @xmath12 of @xmath0 is 2 iff @xmath1 is bipartite and is @xmath15 else . therefore , @xmath16 quantifies the deviation of @xmath1 from being bipartite , and we want to understand this aspect in more detail . in more general terms , we are asking for a quantitative connection between the geometry ( of the graph @xmath1 ) and the analysis ( of the operator @xmath0 , or the random walk encoded by it ) . now , such connections have been explored systematically in riemannian geometry , and many eigenvalue estimates are known there that connect the corresponding laplace operator with the geometry of the underlying space @xmath17 , see e.g. li - yau @xcite , chavel @xcite . the crucial role here is played by the ricci curvature of @xmath17 . in recent years , a kind of axiomatic approach to curvature has been developed . this approach encodes the abstract formal properties of curvature and thereby makes the notion extendible to spaces more general than riemannian manifolds . by now , there exist many notions of generalized curvature , and several of them have found important applications , see sturm @xcite , lott - villani @xcite , ollivier @xcite , ohta @xcite , bonciocat - sturm @xcite , joulin - ollivier @xcite and the references therein . the curvature notion that turns out to be most useful for our purposes is the one introduced by ollivier @xcite . one of our main results says that @xmath18 where @xmath19 is a lower bound for ollivier s ricci curvature . this matches quite well with ollivier s result that the smallest nonzero eigenvalue @xmath11 satisfies @xmath20 in fact , one of the main points of the present paper is to relate such upper and lower bounds via random walks . in fact , as in bauer - jost @xcite , we translate this relationship into the geometric concept of a neighborhood graph . the idea here is that in the @xmath9-th neighborhood graph @xmath4 $ ] of @xmath1 , vertices @xmath14 and @xmath21 are connected by an edge with a weight given by the probability that a random walker starting at @xmath14 reaches @xmath21 after @xmath9 steps times the degree of @xmath14 . we note that even though the original graph may have been unweighted , the neighborhood graphs @xmath4 $ ] are necessarily weighted . in addition , they will in general possess self - loops , because the random walker starting at @xmath14 may return to @xmath14 after @xmath9 steps . therefore , we need to develop our theory on weighted graphs with self - loops even though the original @xmath1 might have been unweighted and without such loops . since ollivier s curvature is defined in terms of transportation distances ( wasserstein metrics ) , we can then use our neighborhood graphs in order to geometrically control the transportation costs and thereby to estimate the curvature of the neighborhood graphs in terms of the curvature of the original graph . as it turns out that lower bounds for the smallest eigenvalue of @xmath22 $ ] are related to upper bounds for the largest eigenvalue of @xmath1 , we can then use methods for controlling the smallest eigenvalue to also obtain ( [ 0a ] ) . for controlling the smallest eigenvalue , besides ollivier @xcite , we also refer to lin - yau @xcite and jost - liu @xcite . in particular , in the last paper , we could relate @xmath11 to the local clustering coefficient introduced in watts - strogatz @xcite . the local clustering coefficients measures the relative local frequency of triangles , that is , cycles of length 3 . since bipartite graphs can not possess any triangles , this then is obviously related to our question about quantifying the deviation of the given graph @xmath1 from being bipartite . in fact , in jost - liu @xcite , this local clustering has been controlled in terms of ollivier s ricci curvature . thus , in the present paper we are closing the loop between the geometric properties of a graph @xmath1 , the spectrum of its graph laplacian , random walks on @xmath1 , and the generalized curvature of @xmath1 , drawing upon deep ideas and concepts originally developed in riemannian geometry and the theory of stochastic processes . in this paper , @xmath23 will denote an undirected , weighted , connected , finite graph of @xmath24 vertices . we do not exclude loops , i.e. , we permit the existence of an edge between a vertex and itself . @xmath25 denotes the set of vertices and @xmath26 denotes the set of edges . if two vertices @xmath27 are connected by an edge , we say @xmath14 and @xmath21 are neighbors , in symbols @xmath28 . the associated weight function @xmath29 : @xmath30 satisfies @xmath31 ( because the graph is undirected ) and we assume @xmath32 whenever @xmath28 and @xmath33 iff @xmath34 ; thus , we do not permit negative weights . for a vertex @xmath35 , its degree @xmath36 is defined as @xmath37 . if @xmath38 whenever @xmath28 , we shall call the graph an unweighted one . we will also speak of a locally finite graph @xmath39 , which is an undirected , weighted , connected graph and satisfies the property that for every @xmath40 , the number of edges connected to @xmath14 is finite . in this subsection , we recall the definition of the normalized graph laplace operator and state some of its basic properties . in particular , we will emphasize the relations between eigenvalues of the laplace operator and random walks on graphs . let @xmath41 denote the space of all real - valued functions on the set @xmath25 . we attach to each vertex @xmath35 a probability measure @xmath42 and then can define the laplace operator . the laplace operator @xmath0 is defined as @xmath43 the measure @xmath42 can also be considered as the distribution of a @xmath44-step random walk starting from @xmath14 . we will choose @xmath45 in the following . note that @xmath46 is possible when @xmath14 has a loop . with this family of probability measures @xmath47 , @xmath0 is just the normalized graph laplace operator studied in grigoryan @xcite and bauer - jost @xcite and is unitarily equivalent to the laplace operator studied in chung @xcite . on a graph @xmath1 without loops , we can also consider a lazy random walk . a lazy random walk is a random walk that does not move with a certain probability , i.e. for some @xmath14 we might have @xmath48 . in this case , the lazy random walk on @xmath1 is equivalent to the usual random walk on the graph @xmath49 that is obtained from @xmath1 by adding for every vertex @xmath14 a loop with a weight @xmath50 . we also have a natural measure @xmath51 on the whole set @xmath25 , @xmath52 which gives an inner product structure on @xmath41 . the inner product of two functions @xmath53 is defined as @xmath54 then @xmath41 becomes a hilbert space , and we can write @xmath55 . by the definition of the degree and the symmetry of the weight function , we can check that * @xmath51 is invariant w.r.t . @xmath47 , i.e. @xmath56 , @xmath57 ; * @xmath51 is reversible w.r.t . @xmath47 , i.e. @xmath58 , @xmath59 . these two facts imply immediately that the operator @xmath0 is nonpositive and self - adjoint on the space @xmath60 . we call @xmath61 an eigenvalue of @xmath0 if there exists some @xmath62 such that @xmath63using this convention it follows from the observation that @xmath0 is self - adjoint and nonpositive that all its eigenvalues are real and nonnegative . in fact , it s well known that ( see e.g. chung @xcite ) @xmath64since our graph is connected we actually have @xmath65 . in chung @xcite it is shown , by proving a discrete version of the cheeger inequality , that @xmath11 is a measure for how easy / difficult it is to cut the graph into two large pieces . furthermore , @xmath66 if and only if @xmath1 is bipartite . in bauer - jost @xcite it is shown , by establishing a dual version of the cheeger inequality for the largest eigenvalue , that @xmath12 is a measure of the size of a locally bipartite subgraph that is connected by only a few edges to the rest in the graph . in the following , we will call @xmath11 the first eigenvalue and @xmath12 the largest eigenvalue of the operator @xmath0 . in this subsection , we recall the neighborhood graph method developed by bauer - jost @xcite . as discussed above , the laplace operator underlies random walks on graphs . in this section , we discuss the deep relationship between eigenvalues estimates for the laplace operator @xmath0 and random walks on the graph @xmath1 by using neighborhood graphs . we first introduce the following notation . for a probability measure @xmath51 , we denote @xmath67 let @xmath68 be the dirac measure at @xmath14 , then we can write @xmath69 . therefore the distribution of a @xmath9-step random walk starting from @xmath14 with a transition probability @xmath70 is @xmath71for @xmath72 . in particular , using the measure ( [ 2 ] ) , the probability that the random walk starting at @xmath14 moves to @xmath21 in @xmath9 steps is given by @xmath73 the idea is now to define a family of graphs @xmath4 $ ] , @xmath74 that encodes the transition probabilities of the @xmath9-step random walks on the graph @xmath1 . the neighborhood graph @xmath4=(v , e[t])$ ] of the graph @xmath75 of order @xmath76 has the same vertex set as @xmath1 and the weights of the edges of @xmath4 $ ] are defined by the transitions probabilities of the @xmath9-step random walk , @xmath77 : = \delta_xp^t(y)d_x.\ ] ] in particular , @xmath78 $ ] and @xmath28 in @xmath4 $ ] if and only if there exists a path of length @xmath9 between @xmath14 and @xmath21 in @xmath1 . we note here that the neighborhood graph method is related to the discrete heat kernel @xmath79 on graphs . for more details about the discrete heat kernel see for instance grigoryan @xcite . we have @xmath80}{d_xd_y}.\ ] ] [ ex1]we consider the following two examples . note that the neighborhood graph @xmath22 $ ] is disconnected . in fact the next lemma shows that this is the case because @xmath1 is bipartite . note furthermore that @xmath81)$ ] . for this example we have @xmath82)$ ] . these examples shows that the neighborhood graph @xmath4 $ ] is in general a weighted graph with loops , even if the original graph @xmath1 is an unweighted , simple graph . [ l4]the neighborhood graph @xmath4 $ ] has the following properties ( see bauer - jost @xcite ) : * if @xmath9 is even , then @xmath4 $ ] is connected if and only if @xmath1 is not bipartite . furthermore , if @xmath9 is even , @xmath4 $ ] can not be bipartite . * if @xmath9 is odd , then @xmath4 $ ] is always connected and @xmath4 $ ] is bipartite iff @xmath1 is bipartite . * @xmath83 = d_x$ ] for all @xmath35 . note that @xmath84 implies that @xmath85)$ ] for all @xmath9 . in particular , we have the same inner product for all neighborhood graphs @xmath4 $ ] . the crucial observation is the next theorem . [ t4 ] the laplace operator @xmath0 on @xmath1 and the laplace operator @xmath86 $ ] on @xmath4 $ ] are related to each other by the following identity : @xmath87.\ ] ] the nonlinear relation between @xmath0 and @xmath86 $ ] leads directly to the following eigenvalue estimates . [ t3 ] let @xmath88 $ ] be a lower bound for the eigenvalue @xmath89 $ ] of @xmath86 $ ] , i.e. , @xmath89\geq \mathcal{a}[t]$ ] . then @xmath90)^{\frac{1}{t}}\leq\lambda_1\leq\cdots\leq\lambda_{n-1}\leq 1+(1-\mathcal{a}[t])^{\frac{1}{t}}\ ] ] if @xmath9 is even and @xmath91)^{\frac{1}{t}}\leq\lambda_1\ ] ] if @xmath9 is odd . [ t5 ] let @xmath92 $ ] be an upper bound for the largest eigenvalue @xmath93 $ ] of @xmath86 $ ] , i.e. @xmath93\leq \mathcal{b}[t]$ ] . ( since @xmath93\leq 1 $ ] for @xmath9 even , we can assume in this case w.l.o.g . that @xmath92\leq 1 $ ] in this case . ) then all eigenvalues of @xmath0 are contained in the union of the intervals @xmath94)^\frac{1}{t}\right ] \bigcup \left[1+(1-\mathcal{b}[t])^\frac{1}{t } , 2\right]\ ] ] if @xmath9 is even and @xmath95)^\frac{1}{t}\ ] ] if @xmath9 is odd . these theorems show how random walks on graphs ( or equivalently neighborhood graphs ) can be used to estimate eigenvalues of the laplace operator . in the rest of this paper we will use these insights to derive lower bounds for @xmath11 and upper bounds for @xmath12 in terms of the ollivier - ricci curvature of a graph . we define a metric @xmath96 on the set of vertices @xmath25 as follows . for two distinct points @xmath27 , @xmath97 is the number of edges in the shortest path connecting @xmath14 and @xmath21 . then , including the family of probability measures @xmath98 , we have a structure @xmath99 , on which the definition of ricci curvature proposed by ollivier @xcite can be stated . for any two distinct points @xmath27 , the ( ollivier- ) ricci curvature of @xmath99 along @xmath100 is defined as @xmath101 here , @xmath102 is the transportation distance between the two probability measures @xmath70 and @xmath103 , in a formula , @xmath104 where @xmath105 is the set of probability measures @xmath106 that satisfy @xmath107 the conditions ( [ 9 ] ) simply ensure that we start with the measure @xmath70 and end up with @xmath103 . intuitively , @xmath102 is the minimal cost to transport the mass of @xmath70 to that of @xmath103 with the distance as the cost function . we also call such a @xmath106 a transfer plan between @xmath70 and @xmath103 , or a coupling of two random walks governed by @xmath70 and @xmath103 respectively . those @xmath106 ( @xmath106 might not be unique ) which attain the infimum value in ( [ 9 ] ) , are called optimal couplings . the optimal coupling exists in a very general setting . for locally finite graphs the existence follows from a simple and interesting argument in remark 14.2 in @xcite . there is a kantorovich duality formula for transportation distances , @xmath108,\ ] ] where @xmath109 for more details about this concept , we refer to villani @xcite , and evans @xcite . applying this definition of curvature to a riemannian manifold with a family of probability measures obtained by restricting the volume measure to the closed @xmath110-balls , @xmath111 will reduce to the ricci curvature as @xmath112 up to a scaling @xmath113 , see ollivier @xcite , for which we also refer for a general treatment of ricci curvature . in riemannian geometry , ricci curvature controls how fast the geodesics emanating from the same point diverge on average , or equivalently how fast the volume of distance balls grows as a function of the radius . jost - liu @xcite translate those ideas into a combinatorial setting and show that ollivier - ricci curvature on a locally finite graph reflects the relative abundance of triangles , which is captured by the local clustering coefficient introduced by watts - strogatz @xcite . for the rest of this paper , let @xmath19 be a lower bound for the ollivier - ricci curvature , i.e. @xmath114 by proposition 19 in ollivier @xcite , it follows that if @xmath19 is a lower curvature bound for all neighbors @xmath115 then it is a lower curvature bound for all pairs of vertices . this also follows from theorem [ bd ] below . by definition , the lower bound @xmath19 for the curvature @xmath111 is no larger than one . in fact , such a lower bound @xmath19 always exists . since the largest possible distance between points from the supports of two measures @xmath70 and @xmath103 at a pair of neighbors @xmath116 is @xmath117 , we can easily estimate @xmath118 . we will derive a more precise lower bound for @xmath111 on a locally finite graph with loops in theorem [ t6 ] , see also lin - yau @xcite and jost - liu @xcite for related results . we could also write ( [ 11 ] ) as @xmath119 which is essentially equivalent to the well known path coupling criterion on the state space of markov chains used to study the mixing time of them ( see bubley - dyer @xcite or @xcite , @xcite ) . we will utilize this idea to interpret the lower bound of the ollivier - ricci curvature as a control on the expectation value of the distance between two coupled random walks . we reformulate bubley - dyer s theorem ( see @xcite or @xcite , @xcite ) in our language . [ bd ] on @xmath99 , if for each pair of neighbors @xmath27 , we have the contraction @xmath120 then for any two probability measures @xmath51 and @xmath121 on @xmath25 , we have @xmath122 with this at hand , it is easy to see that if for any pair of neighbors @xmath116 , @xmath123 , then for any time @xmath9 and any two @xmath124 , which are not necessarily neighbors , the following is true , @xmath125 we consider two coupled discrete time random walks @xmath126 , whose distributions are @xmath127 , @xmath128 respectively . they are coupled in a way that the probability @xmath129 where @xmath130 is the optimal coupling of @xmath127 and @xmath128 . in this language , we can interpret the term @xmath131 as the expectation value of the distance @xmath132 between the coupled random walks @xmath133 and @xmath134 . [ c1 ] on @xmath99 , if @xmath135 , then we have for any two @xmath136 , @xmath137 in his paper @xcite , ollivier proved a spectral gap estimate which works on a general metric space with random walks . in particular , on finite graphs , it could be stated as follows . [ t1 ] on @xmath99 , if @xmath135 , then the first eigenvalue @xmath11 of the normalized graph laplace operator @xmath0 satisfies @xmath138 this is a discrete analogue of the estimate for the smallest nonzero eigenvalue of the laplace - beltrami operator on a riemannian manifold by lichnerowicz . as pointed out in ollivier @xcite , this result is also related to the coupling method for estimates of the first eigenvalue in the riemannian setting developed by chen - wang @xcite ( which leads to a refinement of the eigenvalue estimate of li - yau @xcite ) , see also the surveys chen @xcite . the corresponding result of corollary [ c1 ] in the smooth case , i.e. , controlling the expectation distance of two coupled markov chains in terms of the lower bound of ricci curvature on a riemannian manifold , is a key step in chen - wang s method . a direct proof of theorem [ t1 ] can be found in @xcite . here we want to present an analogue of chen - wang s method in the discrete setting . our proof of theorem [ t1 ] motivated us to combine the ollivier - ricci curvature and the neighborhood graph method via random walks . it reflects the deep connection between eigenvalue estimates and random walks or heat equations . ` proof : ` we consider the transition probability operator @xmath139 defined by @xmath140 . then we have @xmath141 we construct a discrete time heat equation , @xmath142 where @xmath143 satisfies @xmath144 . iteratively , one can find the solution of the above system of equations as @xmath145 we remark here that the solution of the heat equation on a riemannian manifold with the first eigenfunction as the initial value is @xmath146 , which also involves information about both the eigenvalue @xmath11 and the eigenfunction @xmath143 . if @xmath147 there is nothing to prove since @xmath148 by definition . therefore we can suppose @xmath149 in the following . then we have for any @xmath136 @xmath150 here , @xmath151 is always finite since the underlying space @xmath25 is a finite set . in the last inequality we used corollary [ c1 ] . from an analytic point of view , the above calculation can be seen as a gradient estimate for the solution of the heat equation . since the eigenfunction @xmath152 for the eigenvalue @xmath11 is orthonogal to the constant function , i.e. @xmath153 , we can always find @xmath154 such that @xmath155 . it follows that @xmath156 to prevent a contradiction when @xmath157 , we need @xmath158 which completes the proof . jost - liu @xcite give a sharp estimate for ollivier - ricci curvature on locally finite graphs without loops . as shown in example [ ex1 ] , neighborhood graphs are in general weighted graphs with loops . therefore , for our purposes , we need to understand the curvature of graphs with loops . in this section , we generalize the estimates in jost - liu @xcite for locally finite graphs @xmath160 that may have loops . this is done by considering a novel optimal transportation plan . we first fix some notations . for any two real numbers @xmath161 , @xmath162 we denote @xmath163 as the neighborhood of @xmath14 and @xmath164 . then @xmath165 if @xmath14 has a loop . for every pair of neighbors @xmath166 , we divide @xmath167 into disjoint parts as follows . @xmath168 where @xmath169 and @xmath170 in the next figure we illustrate this partition of the vertex set . [ t6 ] on @xmath160 , we have for any pair of neighbors @xmath171 , @xmath172 moreover , this inequality is sharp . [ re2 ] on an unweighted graph , the form @xmath173 for @xmath28 becomes @xmath174 where @xmath175 is the number of triangles containing @xmath115 , @xmath176 or @xmath44 is the number of loops at @xmath14 . ` proof : ` since the total mass of @xmath70 is equal to one , we obtain from ( [ eq3 ] ) the following identity for neighboring vertices @xmath14 and @xmath21 : @xmath177 a similar identity holds for @xmath21 . we denote @xmath178 obviously , @xmath179 . we firstly try to understand these two quantities . if @xmath180 , we have @xmath181 i.e. , using ( [ eq4 ] ) we observe that the mass of @xmath103 at @xmath21 and @xmath182 is no smaller than that of @xmath70 at @xmath21 and the excess mass at @xmath183 . rewriting ( [ 31 ] ) in the form @xmath184and subtracting the term @xmath185 on both sides we obtain @xmath186 i.e. , the mass of @xmath70 at @xmath14 and @xmath187 is larger than that of @xmath103 at @xmath14 and the excess mass at @xmath188 . if @xmath189 , we have @xmath190 i.e. , the mass of @xmath70 at @xmath14 and @xmath187 and the excess mass at @xmath183 is no smaller than that of @xmath103 at @xmath14 . in jost - liu @xcite it is explicitly described how much mass has to be moved from a vertex in @xmath191 to which point in @xmath192 , i.e. the exact value of @xmath193 , for any @xmath194 , @xmath195 . but in the case with loops it would be too complicated if we try to do the same thing . instead , we adopt here a dynamic strategy . that is , we think of a discrete time flow of mass . after one unit time , the mass flows forward for distance @xmath44 or stays there . we only need to determine the direction of the flow according to different cases . as in jost - liu @xcite , we divide the discussion into 3 cases . * @xmath196 in this case we use the following transport plan : suppose the initial time is @xmath197 . + @xmath6 : : move all the mass at @xmath187 to @xmath14 and the excess mass at @xmath183 to @xmath21 . we denote the distribution of the mass after the first time step by @xmath198 . we have @xmath199 @xmath200 : : move one part of the excess mass at @xmath14 now to fill the gap at @xmath188 and the other part to @xmath21 . by ( [ 32 ] ) the mass at @xmath14 after @xmath6 is enough to do so . the distribution of the mass is now denoted by @xmath201 . we have @xmath202\times 1\end{aligned}\ ] ] @xmath203 : : move the excess mass at @xmath21 now to @xmath182 . we denote the mass after the third time step by @xmath204 . we have @xmath205\times 1\end{aligned}\ ] ] + by triangle inequality and ( [ 9 ] ) , we get @xmath206 moreover , if the following function can be extended as a function on the graph such that @xmath207 , ( i.e. , if there are no paths of length 1 between @xmath187 and @xmath188 , nor paths of length 1 between @xmath182 and @xmath183 , nor paths of length @xmath44 or @xmath208 between @xmath187 and @xmath182 , ) @xmath209 then by kantorovich duality ( [ 10 ] ) , we can show that the inequality above is actually an equality . recalling the definition of @xmath210 , we have proved the theorem in this case . * @xmath211 we use the following transfer plan : + @xmath6 : : we divide the excess mass of @xmath70 at @xmath183 into two parts . one part together with the mass of @xmath70 at @xmath21 is enough to fill gaps at @xmath21 and @xmath182 . since ( [ 31 ] ) does nt hold in this case , this is possible . we move this part of mass to @xmath21 and the other part to @xmath14 . we also move all the mass of @xmath70 at @xmath212 to x. @xmath200 : : we move the excess mass at @xmath14 now to @xmath188 and the excess mass at @xmath21 to @xmath213 . + applying this transfer plan , we can prove ( we omit the calculation here ) @xmath214 moreover , if the following function can be extended as a function on the graph such that @xmath207 , ( i.e. , if there are no paths of length 1 between @xmath215 and @xmath216 , ) @xmath217 then by kantorovich duality ( [ 10 ] ) , we can check that the inequality above is actually an equality . * @xmath218 we use the following transport plan : + @xmath6 : : move the mass of @xmath70 at @xmath187 and @xmath183 to @xmath14 . since now ( [ 33 ] ) does nt hold , we need to move one part of the mass @xmath219 to @xmath14 and the other part to @xmath182 and @xmath188 . + applying this transfer plan , we can calculate @xmath220 since the following function can be extended as a function on the graph such that @xmath207 , @xmath221 we can check the inequality above is in fact an equality by kantorovich duality . that is , in this case for any @xmath28 , @xmath222 @xmath159 we also have a generalization of the upper bound in jost - liu @xcite on @xmath223 . [ t7 ] on @xmath160 , we have for every pair of neighbors @xmath116 , @xmath224 ` proof : ` @xmath225 is exactly the mass of @xmath70 which we need not move . the other mass need to be moved for at least distance 1 . so we have @xmath226 , which implies @xmath227 , for @xmath28 . @xmath159 if for every @xmath40 , @xmath228 , then theorem [ t6 ] and theorem [ t7 ] reduce to the estimates in jost - liu @xcite . [ re1 ] theorem [ t7 ] tells us that if @xmath229 for any @xmath230 , then either @xmath231 or @xmath232 or @xmath233 . [ ex2 ] we consider a lazy random walk on an unweighted complete graph @xmath234 with @xmath24 vertices governed by @xmath235 . or equivalently , we consider the graph @xmath236 . using theorem [ t6 ] and theorem [ t7 ] , we get for any @xmath116 @xmath237 that is , in this case , both the lower and the upper bound are sharp . in this section , we use the neighborhood graph method to derive estimates for the spectrum of the normalized graph laplace operator in terms of the ollivier - ricci curvature . in particular , we get a sharp upper bound for the largest eigenvalue using the curvature on the original graph . furthermore , we give nontrivial estimates using the curvature on neighborhood graphs even if the curvature of the original graph is nonpositive . in this case , our results improve theorem [ t1 ] . [ t8 ] let @xmath3 $ ] be a lower bound of ollivier - ricci curvature of the neighborhood graph @xmath4 $ ] . then for all @xmath76 the eigenvalues of @xmath0 on @xmath1 satisfy @xmath238)^{\frac{1}{t}}\leq \lambda_1 \leq \cdots \leq \lambda_{n-1}\leq 1 + ( 1- k[t])^{\frac{1}{t}}\ ] ] if @xmath9 is even and @xmath239)^{\frac{1}{t}}\leq \lambda_1\ ] ] if @xmath9 is odd . moreover , if @xmath1 is not bipartite , then there exists a @xmath240 such that for all @xmath241 the eigenvalues of @xmath0 on @xmath1 satisfy @xmath242)^{\frac{1}{t}}\leq \lambda_1 \leq \cdots \leq \lambda_{n-1}\leq 1 + ( 1- k[t])^{\frac{1}{t}}<2\]]if @xmath9 is even and @xmath242)^{\frac{1}{t}}\leq \lambda_1\ ] ] if @xmath9 is odd . ` proof : ` by combining theorem [ t3 ] and theorem [ t1 ] we directly obtain equation ( [ eq1 ] ) and equation ( [ eq2 ] ) . the second part of this lemma is proved in two steps . in the first step , we will show that if @xmath1 is not bipartite then there exists a @xmath243 such that for all @xmath244 the neighborhood graph @xmath4 $ ] of @xmath1 satisfies @xmath245\neq 0 $ ] for all @xmath246 , i.e. @xmath4 $ ] is a complete graph and each vertex has a loop . in the second step , we show that any graph that satisfies @xmath247 for all @xmath248 has a positive lower curvature bound , i.e. @xmath249 . this then completes the proof . * step 1 : * by the definition of the neighborhood graph it is sufficient to show that for all @xmath244 there exists a path of length @xmath9 between any pair of vertices . since @xmath1 is not bipartite it follows that there exists a path of even and a path of odd length between any pair of vertices in the graph ( this is the definition of a bipartite graph ! ) . given a path of length @xmath250 between @xmath14 and @xmath21 then we can find a path of length @xmath251 between @xmath14 and @xmath21 as follows : we go in @xmath250 steps from @xmath14 to @xmath21 and then from @xmath21 to one of its neighbors and then back to @xmath21 . this is a path of length @xmath251 between @xmath14 and @xmath21 . since @xmath1 is finite , it follows that there exists a @xmath243 such that there are paths of length @xmath9 for all @xmath244 for any pair of vertices . * step 2 : * given a graph that satisfies @xmath252 for all @xmath246 . since each vertex in the graph is a neighbor of all other vertices , it is clear that we can move the excess mass of @xmath70 for distance @xmath44 to anywhere . therefore @xmath253 which implies @xmath254 by theorem [ t7 ] , it follows that the above inequality is in fact an equality . hence for all @xmath27 , we have @xmath255 since the weight is positive for every pair of vertices . this completes the proof . @xmath159 from remark [ re1 ] after theorem [ t7 ] we know that positive lower curvature bound is a strong restriction on a graph . hence , olliver s estimate @xmath256 only yields nontrivial estimates for @xmath11 in those restricted cases . lemma [ t8 ] shows that , unless @xmath1 is bipartite , the neighborhood graph method can be used to always obtain a nontrivial lower bound for @xmath11 in terms of the ollivier - ricci curvature . now , we will show that it is also possible to control the largest eigenvalue of @xmath0 on the graph @xmath1 in terms of the olliver - ricci curvature of @xmath1 itself . [ t2 ] on @xmath99 , if @xmath135 , then the largest eigenvalue @xmath12 of the normalized graph laplace operator @xmath0 satisfies @xmath257 as discussed above , the largest eigenvalue @xmath12 is small if the graph does not contain a locally bipartite subgraph that has only a few connections to the rest of the graph . it is well known that a graph is bipartite if and only if it does not contain any cycle of odd length , in particular a bipartite graph does not contain triangles and loops . since positive ollivier - ricci curvature implies that locally there always exists a triangle or a loop , i.e. the graph can not be locally bipartite , it is not surprising that the largest eigenvalue can actually be controlled from above in terms of the ollivier - ricci curvature . by combining theorem [ t2 ] and theorem [ t6 ] , we have the following corollary . from remark [ re2 ] , we know that @xmath258 reflects the number of triangles and loops in the graph . on @xmath23 , the largest eigenvalue satisfies @xmath259 where @xmath173 is defined in theorem [ t6 ] . the main point in the proof of theorem [ t2 ] is to explore the relation of the ollivier - ricci curvature @xmath111 on the original graph @xmath1 and the ollivier - ricci curvature @xmath260 $ ] on its neighborhood graph @xmath4 $ ] . before we prove theorem [ t2 ] we consider some lemmata . if we interpret the graph @xmath23 as a structure @xmath261 , then by ( [ 4 ] ) its neighborhood graph @xmath4 = ( v , e[t])$ ] can be considered as a structure @xmath262 , \{\delta_x p^t\})$ ] . so the first step should be to estimate the variance of the metrics on neighborhood graphs . [ l1 ] for any @xmath27 , we have @xmath263(x , y).\ ] ] ` proof : ` for any @xmath246 , we set @xmath264(x , y ) = \infty$ ] if we can not find a path connecting them in @xmath4 $ ] . otherwise , we just choose a shortest path @xmath265 , between @xmath14 and @xmath21 in @xmath4 $ ] , i.e. @xmath266(x , y)$ ] . for @xmath267 , @xmath268 , by definition of neighborhood graph , we have @xmath269 in @xmath1 . equivalently , @xmath270(x_i , x_{i+1}).\ ] ] summing over all @xmath271 , we get @xmath272(x , y).\ ] ] then the triangle inequality of @xmath96 on @xmath1 gives ( [ 18 ] ) . @xmath159 in fact , when @xmath9 is larger than the diameter @xmath273 of the graph @xmath1 , we have a better estimate @xmath274(x , y).\ ] ] [ l2 ] if @xmath275 $ ] , then @xmath264(x , y)\leq d(x , y)$ ] . the proof is obvious . the interesting point is that when the ollivier - ricci curvature of the graph @xmath1 is positive , @xmath276 $ ] is satisfied for all @xmath9 and hence lemma [ l2 ] is applicable . this can be seen as follows . first we observe that for @xmath277 , if @xmath231 or @xmath232 or @xmath233 , then @xmath278 $ ] . then we get the conclusion immediately from remark [ re1 ] after theorem [ t7 ] . [ l3 ] let @xmath19 be a lower bound of @xmath111 on @xmath1 . if @xmath275 $ ] , then the curvature @xmath260 $ ] of the neighborhood graph @xmath4 $ ] satisfies @xmath279(x , y)\geq 1-t(1-k)^t , \,\,\,\forall x , y\in v.\ ] ] ` proof : ` by lemma [ l2 ] , corollary [ c1 ] and lemma [ l1 ] , we get @xmath280}(\delta_x p^t , \delta_y p^t)&\leq w_1^d(\delta_x p^t , \delta_y p^t)\\ & \leq ( 1-k)^td(x , y)\\&\leq t(1-k)^td[t](x , y).\end{aligned}\ ] ] we use @xmath281}$ ] , @xmath282 here to indicate the different cost functions used in these two quantities . in the first inequality above we used that the transportation distance ( [ 9 ] ) is linear in the graph distance @xmath283 . recalling the definition of the curvature , we have proved ( [ 21 ] ) . @xmath159 now we arrive at the point to prove the upper bound of the largest eigenvalue . ` proof of theorem ` [ t2 ] : using lemma [ l3 ] and theorem [ t1 ] , we know on @xmath4 $ ] , @xmath284\geq 1-t(1-k)^t.\ ] ] then by lemma [ t8 ] , we get for any even number @xmath9 , @xmath285 letting @xmath286 , we get @xmath8 . @xmath159 by combining theorem [ t1 ] and theorem [ t2 ] , we get @xmath287 that is , ( [ eq1 ] ) is also true for @xmath6 . [ e1 ] on an unweighted complete graph @xmath288 with @xmath24 vertices , we have @xmath289 therefore , @xmath290 that is , our upper bound estimate for @xmath12 is sharp for unweighted complete graphs . let s revisit the graph @xmath236 in example [ ex2 ] . we have @xmath291 therefore , @xmath292 that is , both estimates are sharp in this case . in fact , we can show that ( [ eq1 ] ) is true for any @xmath76 . [ t10 ] let @xmath3 $ ] be a lower bound of ollivier - ricci curvature of the neighborhood graph @xmath4 $ ] . then for all @xmath76 the eigenvalues of @xmath0 on @xmath1 satisfy @xmath2)^{\frac{1}{t}}\leq \lambda_1 \leq \cdots \leq \lambda_{n-1}\leq 1 + ( 1- k[t])^{\frac{1}{t}}.\ ] ] moreover , if @xmath1 is not bipartite , then there exists a @xmath240 such that for all @xmath244 the eigenvalues of @xmath0 on @xmath1 satisfy @xmath242)^{\frac{1}{t}}\leq \lambda_1 \leq \cdots \leq \lambda_{n-1}\leq 1 + ( 1- k[t])^{\frac{1}{t}}<2.\ ] ] ` proof : ` by lemma [ t8 ] , we only need to prove the upper estimate @xmath293)^{\frac{1}{t}}$ ] if @xmath9 is odd . this follows immediately from theorem [ t2 ] and theorem [ t5].@xmath159 in particular , unless @xmath1 is bipartite , we can obtain nontrivial lower bounds for @xmath11 and nontrivial upper bounds for @xmath12 even if @xmath294 $ ] is nonpositive . in bauer - jost @xcite it is shown that the next lemma is a simple consequence of theorem [ t4 ] . [ l5 ] let @xmath295 be an eigenfunction of @xmath0 for the eigenvalue @xmath61 . then , @xmath296u)_{\mu}}{(u , \delta u)_{\mu}}=\frac{\sum_{x , y } w_{xy}[2](u(x)-u(y))^2}{\sum_{x , y } w_{xy}(u(x)-u(y))^2}.\ ] ] lemma [ l5 ] can be used to derive further estimates for the largest eigenvalue @xmath12 form above and below . we introduce the following notations : [ 69 ] let @xmath297 be the neighborhood of vertex @xmath14 as in section [ section3 ] . the minimal and the maximal number of joint neighbors of any two neighboring vertices is defined as @xmath298 and @xmath299 , respectively . furthermore , we define @xmath300 and @xmath301 . [ 65]we have the following estimates for @xmath12 : * if @xmath302)$ ] then @xmath303 * if @xmath304)\subseteq e(g)$ ] then @xmath305 on the one hand , we observe that if @xmath302)$ ] then for every pair of neighboring vertices @xmath28 in @xmath1 @xmath306}{w_{xy}}=\frac{\sum_{z}\frac{1}{d_{z}}w_{xz}w_{zy}}{w_{xy } } & \geq & \frac{w^2}{w}\frac{\tilde{\sharp}_1}{\max_x d_x}.\end{aligned}\ ] ] on the other hand if @xmath304)\subseteq e(g)$ ] then for every pair of neighboring vertices @xmath28 in @xmath307 we have @xmath308}{w_{xy}}=\frac{\sum_{z}\frac{1}{d_{z}}w_{xz}w_{zy}}{w_{xy}}\leq \frac{w^2}{w}\frac{\tilde{\sharp}_2}{\min_x d_x}.\ ] ] substituting the inequalities ( [ 64 ] ) and ( [ 63 ] ) in equation ( [ highest2 ] ) completes the proof . interestingly , theorem [ 65 ] @xmath309 yields an alternative proof of theorem [ t2 ] in the case of unweighted regular graphs . since theorem [ t2 ] trivially holds if @xmath13 we only have to consider the case when @xmath249 is a lower curvature bound . the discussion after lemma [ l2 ] shows that @xmath249 implies that @xmath310)$ ] and hence we can apply theorem [ 65 ] ( i ) in this case . from theorem [ t7 ] it follows that for an unweighted graph @xmath311 for all pairs or neighboring vertices @xmath115 . in the case of an @xmath96-regular graph @xmath1 this implies that a lower bound @xmath19 for the ollivier - ricci curvature must satisfy , @xmath312 hence for an unweighted @xmath96-regular graph theorem [ 65 ] implies @xmath313 we consider the following example . let s revisit the complete unweighted graph @xmath234 . for the complete graph we have @xmath314)$ ] . we have @xmath315 and @xmath316 . thus , theorem [ 65 ] @xmath309 yields @xmath317 i. e. the estimate from above is sharp for complete graphs . now we consider the unweighted complete graph on @xmath24 vertices with @xmath24 loops @xmath318 . we have @xmath319 ) = e({\mathcal k}_n^{\mathrm{lazy}})$ ] . furthermore , we have @xmath320 and @xmath321 . thus , theorem [ 65 ] @xmath322 yields @xmath323 i. e. the estimate from below is sharp for complete graphs with loops . in this section , we explore a particular example , the circle @xmath324 with @xmath325 vertices . we show that our estimates using the neighborhood graph method can yield nontrivial estimates although the curvature of the original graph has a nonpositive lower curvature bound . we also discuss the growth rate of the lower bound @xmath19 for curvature @xmath260 $ ] as @xmath326 on @xmath324 . we know that the first and largest eigenvalue of @xmath0 on @xmath324 are given by @xmath327 it is easy to check that the optimal lower bound @xmath19 for the curvature is @xmath328 . so in this case ollivier s result theorem [ t1 ] and theorem [ t2 ] only yield trivial estimates . the weight of every dashed loop is @xmath44 and the weight of every solid edge is @xmath330 . we can check that the optimal @xmath331 $ ] is @xmath332 . then theorem [ t10 ] yields the nontrivial estimate , @xmath333 the weight of every dashed loop is @xmath335 and the weight of every solid edge is @xmath330 and every dash - dotted edge is @xmath336 . we can check that the optimal lower curvature bound is @xmath337= 1/2 $ ] . so theorem [ t10 ] tells us that @xmath338{2}}\doteq 0.1591,\,\,\lambda_4\leq 1+\frac{1}{\sqrt[4]{2}}\doteq 1.8409.\ ] ] the weight of every solid edge is @xmath335 and the weight of every dashed edge is @xmath332 . then the optimal lower curvature bound is given by @xmath340=3/8 $ ] . theorem [ t10 ] implies @xmath341 from the above calculations , we see that the ollivier - ricci curvature @xmath260 $ ] on the neighborhood graphs give better and better estimates in this example . we also further compare the estimates in theorem [ 65 ] and theorem [ t2 ] on the weighted graph @xmath339 $ ] . by theorem [ 65 ] we get @xmath342\leq 2-\frac{(1/4)^2}{3/4}\cdot\frac{3}{2}=\frac{15}{8}.\ ] ] this is worse than the one given by theorem [ t2 ] , @xmath343\leq 2-\frac{3}{8}=\frac{13}{8}$ ] . another interesting problem is concerned with the limit of the neighborhood graphs . as shown in bauer - jost @xcite , since @xmath324 is a regular non - bipartite graph , @xmath344 $ ] will converge to @xmath345:=\bar{\mathcal{c}}_5 $ ] as @xmath326 . in this case @xmath346 is a complete graph and every vertex has a loop and the weight of every edge in @xmath346 is @xmath347 . we can then check that @xmath348=1 $ ] . by theorem [ t10 ] , for large enough @xmath9 , we have @xmath349)^{\frac{1}{t}}\leq\lambda_1=1-\cos\frac{2\pi}{5}<1.\ ] ] therefore @xmath350)^{\frac{1}{t}}<1 $ ] . intuitively , @xmath3 $ ] should become larger and larger as @xmath157 since the graph @xmath344 $ ] converges to a complete graph and its weights become more and more the same . we suppose @xmath351)^{\frac{1}{t}}$ ] exists ( this is true at least for a subsequence ) . then to avoid contradictions in ( [ 22 ] ) , we know there exists a positive number @xmath352 such that @xmath353)^{\frac{1}{t}}=e^{-a}>0.\ ] ] that is , @xmath354)}{t}=-a,\ ] ] which means @xmath3 $ ] behaves like @xmath355 as @xmath356 where @xmath357 is a polynomial in @xmath9 .	we prove the following estimate for the spectrum of the normalized laplace operator @xmath0 on a finite graph @xmath1 , @xmath2)^{\frac{1}{t}}\leq \lambda_1 \leq \cdots \leq \lambda_{n-1}\leq 1 + ( 1- k[t])^{\frac{1}{t } } , \,\forall \,\,\text{integers}\,\ , t\geq 1.\ ] ] here @xmath3 $ ] is a lower bound for the ollivier - ricci curvature on the neighborhood graph @xmath4 $ ] ( here we use the convention @xmath5=g$ ] ) , which was introduced by bauer - jost . in particular , when @xmath6 this is ollivier s estimate @xmath7 and a new sharp upper bound @xmath8 for the largest eigenvalue . furthermore , we prove that for any @xmath1 when @xmath9 is sufficiently large , @xmath10)^{\frac{1}{t}}$ ] which shows that our estimates for @xmath11 and @xmath12 are always nontrivial and the lower estimate for @xmath11 improves ollivier s estimate @xmath7 for all graphs with @xmath13 . by definition neighborhood graphs possess many loops . to understand the ollivier - ricci curvature on neighborhood graphs , we generalize a sharp estimate of the curvature given by jost - liu to graphs which may have loops and relate it to the relative local frequency of triangles and loops .
gravitational microlensing is a unique technique to find exoplanets down to low masses @xcite just beyond the snow line @xcite which plays an important role in the core accretion theory of planet formation @xcite . microlensing is presently the only technique able to find planets down to an earth size mass at these orbital radii . microlensing is also sensitive to planets orbiting around faint stars like m - dwarfs and brown dwarfs , and can even detect free - floating planets @xcite because it does not rely on the host s light . this is complimentary to the other methods like the radial velocity @xcite and the transit @xcite methods which are sensitive to planets close to their host stars . several survey groups are conducting high cadence survey observations for microlensing events towards the galactic bulge . to date , about 50 exoplanets have been found by microlensing thanks to their continuous effort . several statistical studies have revealed the planet abundances beyond the snow line @xcite . @xcite found a possible peak in the mass ratio function and that cold neptunes are likely to be the most common type of planets beyond the snow line . recently , @xcite suggested that there exists a possible paucity of planets in the galactic bulge from the sample of observed microlensing planets , although it appears that they have overestimated the detection efficiency for planets orbiting bulge stars . distances to the planetary system is determined when a microlensing parallax signal or lens star flux are measured from high resolution follow - up observations . most planetary systems which are likely to be located in the galactic bulge have their distance estimated by a bayesian analysis with somewhat uncertain priors because microlensing parallax is not easily detected . one event , moa-2011-blg-293 lb was determined to be located in the galactic bulge by measuring the lens flux @xcite , but @xcite and @xcite have shown that there can be ambiguity in the interpretation of these apparent lens detections . the recent simultaneous observations from the _ spitzer _ space telescope helped to measure the space - based parallax and contribute to the determination of the distance to the lens system @xcite . however the statistical sample of microlensing planets is not yet large enough to draw a clear picture . the wfirst satellite is expected to find up to about 2600 exoplanets by microlensing including 370 earth - mass planets in the mid 2020s @xcite . in 70 % of these events , the masses and the distance will be measured . until then , it is important to increase the number of planetary microlensing events by ground - based telescopes as far as possible . in this paper , we report the analysis of a planetary microlensing event ogle-2013-blg-1761 . in section 2 , we describe the observations of this event and section 3 describes the data reduction . in section 4 , we describe the modeling of the light curve . an analysis of the microlensed source and the einstein radius is given in section 5 . in section 6 , we discuss our likelihood analysis . finally , we discuss the results of this work in section 7 . the microlensing observations in astrophysics collaboration ( moa ; @xcite ) conducts a high cadence microlensing survey observation program towards the galactic bulge at the mt john university observatory in new zealand using the 1.8 m moa - ii telescope equipped with a very wide field - of - view ( 2.2 @xmath8 ) moa - cam3 ccd camera @xcite . moa observes with cadences ranging from 15 to 90 minutes depending on the target field . the moa - ii observations are carried out in the custom moa - red wide band filter , which corresponds to the sum of the standard cousins @xmath9 and @xmath10-bands . moa issues @xmath11 alerts of microlensing events in real time each year . the optical gravitational lensing experiment ( ogle ; @xcite ) conducts a microlensing survey at the las campanas observatory in chile using the 1.3 m warsaw telescope equipped with a 1.4 @xmath8 fov ogle - iv camera . ogle observes the galactic bulge fields with cadences ranging from 20 minutes to once a night . most observations are taken in the standard kron - cousin @xmath10-band with occasional observations in the johnson @xmath12-band . ogle issues alerts for @xmath13 microlensing events in real time each year . the gravitational microlensing event ogle-2013-blg-1761 was first found and alerted by ogle on 4 sep 2013 ( hjd = hjd-2450000 @xmath14 6540 ) . moa independently detected this event on 26 sep ( hjd = 6562 ) as moa-2013-blg-651 . the coordinate of this event is ( r.a . , decl)(j2000 ) = ( @xmath15,@xmath16 ) or @xmath17 in galactic coordinates . the observational cadences for this field are once per hour for ogle and once per 15 minutes for moa , respectively . this area is highly reddened because of interstellar dust . figure [ fig1 ] shows the light curve of this event . on september 30 ( hjd = 6566 ) , the moa collaboration detected an short deviation in the light curve and announced the anomaly . unfortunately , we could not get a good coverage of the anomaly due to the bad weather in the following day for both ogle and moa . but it was enough for determining its nature as is shown in the following analysis . the source was faint and expected not to reach high magnification . besides this event was near the end of the observational season and the observing time for the galactic bulge was limited . thus only the survey groups , moa and ogle observed this event . the moa - red - band data was reduced by the moa difference image analysis ( dia ) pipeline @xcite . systematic errors in the moa photometry were detrended for effects due to the differential refraction and seeing by using photometry taken outside of the magnified part of the light curve . here the psf of the target was affected by a near star when the seeing was large . we also removed data with seeing @xmath18 4.5 pixels ( @xmath19 ) in which the effect is too large to correct . the ogle - i and v - band data were reduced by the ogle dia photometry pipeline @xcite . our modeling light curve comprises 24632 moa - red data points , 4929 ogle - i data points and 119 ogle - v data points . it is known that the nominal errors from the photometric pipeline are underestimated for the stellar dense fields like ours . we renormalized the error bars of each dataset by using the following standard formula presented in @xcite , @xmath20 where @xmath21 is the original error of the @xmath22th data point in magnitudes , and @xmath23 and @xmath24 are the re - normalizing parameter . the cumulative @xmath25 distribution from the tentative best fit model , which are sorted by their magnification of the model at each data point , is supposed to be the straight line if the data follow a normal distribution . thus @xmath24 is chosen to make this cumulative @xmath25 distribution a straight line . then @xmath23 is chosen so that each data set gives @xmath26 . we obtained @xmath27 for all data sets and @xmath28 for the moa - red , @xmath29 and @xmath30 for the ogle - i and v - band , respectively . the data sets we used are listed in table [ tab1 ] . for the point - source point - lens ( pspl ) model , there are three parameters to characterize the microlens light curve , @xmath31 : the time of closest approach of the source to the lens masses , @xmath32 : the minimum impact parameter in units of the angular einstein radius @xmath33 , and @xmath34 : the einstein radius crossing time in day . for the binary lens model , there are three additional parameters , @xmath35 : the planet / host mass ratio , @xmath36 : the projected planet - star separation in units of the einstein radius , and @xmath37 : the angle of the source trajectory relative to the binary lens axis . when we take account of the finite source effect and the parallax effect , the angular radius of the source star in units of @xmath33 , @xmath38 , and the east and north components of the microlensing parallax vector , @xmath39 and @xmath40 , are added for each case . the model light curve can be given by @xmath41 where f(@xmath42 ) is the flux at time _ t _ , a(@xmath42 ) is a magnification of the source star at _ @xmath43 and @xmath44 are baseline fluxes from the source and blend stars , respectively . we use linear limb darkening models for the source star . the effective temperature of the source star estimated from the extinction corrected source color , @xmath45 as discussed in section 5 is @xmath46 @xcite . assuming @xmath47 , surface gravity @xmath48 and metallicity @xmath49 = 0 $ ] , we selected limb darkening coefficients @xmath50 to be 0.6534 , 0.6049 and 0.7796 for moa - red , ogle - i and ogle - v bands , respectively @xcite . the moa - red values is the mean of the _ r_- and _ i_-band values . the limb darkening coefficients are also listed in table [ tab1 ] . we searched for the best - fit model over a wide range of values of microlensing parameters by using the markov chain monte carlo ( mcmc ) algorithm @xcite and the image - centered ray - shooting method @xcite . to find the global best model , we first conduct a grid search by fixing three parameters , @xmath35 , @xmath36 and @xmath37 , at 9680 different grid points with other parameters free . next , by using the best 100 smallest @xmath25 models as a initial parameters , we search for the best - fit model by refining all parameters . the best - fit light curve and the parameters are shown in figure [ fig1 ] and table [ tab2 ] . in the initial grid search , we found two degenerate solutions with @xmath51 ( close model ) and @xmath52 ( wide model ) . for high magnification planetary events , this close - wide " degeneracy often occurs because the shape of the central caustic of close and wide models are very similar to each other @xcite . this is usually not the case for the low magnification planetary events where the source star passes the planetary caustic or a resonant caustic such as in this event . however , the anomaly of this event was only sparsely sampled especially when the source was expected to exit the caustic . the close solution is only slightly preferred by @xmath53 . thus we can not distinguish these two models . the mass ratios are @xmath0 and @xmath1 for close ( @xmath2 ) and wide ( @xmath3 ) models , respectively . this indicates that both models have planetary mass ratios though the uncertainty of @xmath35 is large due to the poor coverage . the @xmath54 of these models compared to the pspl and fspl ( finite - source point - lens ) are about 1401 and 1234 , respectively , so the planetary signal is detected confidently . we also searched for the best - fit model without including finite source effects . in the best - fit model , the source crosses the caustic , but the caustic exit was not observed . so we checked if there are models that explain the light curve without crossing the caustic . we conducted a grid search in the same way except that the value of @xmath38 was fixed to be zero . we found a best - fit model without including the finite source effect which has a planetary mass ratio @xmath55 . the difference in the @xmath25 between the best - fit model with and without finite source effect is @xmath56 . so the finite source effect is detected with @xmath57 confidence . microlensing parallax is an effect where the orbital motion of the earth deviates the apparent lens - source relative motion from a inertial trajectory @xcite . this can be described by the microlensing parallax vector @xmath58 whose direction is the direction of the lens - source relative motion projected on the sky and its amplitude , @xmath59 , is the inverse of the einstein radius , projected to the observer plane . the orbital parallax is likely to be measured when the event time scale is relatively large and the event is observed in the autumn like this event because the acceleration of the earth projected to the bulge becomes biggest at the spring and the autumnal equinoxes . if the parallax effect and finite source effect are measured in a gravitational microlensing event , we can calculate the lens properties uniquely by assuming the distance to the source star , @xmath60 , as @xmath61 and @xmath62 where @xmath63 and @xmath64 @xcite . we fitted a light curve including the effect of parallax with the parameters of the best - fit non - parallax model as initial parameters . our best - fit parallax model has @xmath65 , but the improvement in @xmath25 is only @xmath66 . we checked where the parallax signal originated from by examining the cumulative @xmath54 between the best static model and the model with parallax effect as shown in figure [ fig3 ] . the cumulative @xmath54 for the moa - red , ogle - i and -v are plotted separately . we can see that the moa - red favors the parallax model but the ogle - i does not . they are inconsistent with each other . thus we concluded that the parallax effect was not detected for this event . to check the uniqueness of the best planetary model , we inspected the models with mass ratio @xmath35 in the range of @xmath67 carefully . figure [ fig4 ] shows @xmath54 from the best - fit model as a function of @xmath35 . the dotted vertical line indicates the mass ratio @xmath68 which is the nominal boundary between the planet - star binary star systems . we can see that the @xmath54 between the best model and the model with @xmath69 is larger than @xmath70 . we conclude that the best planetary model is superior to any binary models more than @xmath71 confidence . the source angular radius , @xmath72 , can be derived from the intrinsic source color and the magnitude . combined with the source star radius , @xmath38 , obtained from our light curve modeling including the finite source effect , we can estimate the angular einstein radius @xmath73 . the ogle data were taken in the ogle - iv _ i _ and _ v _ bands . we calibrated them to the standard _ vi _ photometric system . we used the following relations to calibrate the ogle _ vi _ magnitude to the standard _ vi _ magnitude @xcite , @xmath74 @xmath75 the source color and magnitude measured from the light curve fitting are affected by the reddening and extinction due to the interstellar dust . to obtain the intrinsic source color and magnitude , we used the centroid of red clump giants ( rcg ) as standard candle . figure [ fig5 ] shows the ogle - iv calibrated color magnitude diagram ( cmd ) for the stars within @xmath76 around the source star . the centroid of rcg , @xmath77 and the calibrated source color and magnitude obtained from the light curve fitting , @xmath78 are shown as filled red and orange circles , respectively . the error in @xmath79 includes the errors from the mcmc , the linear fit of the light curve fitting and the uncertainty in the calibration relations . in figure [ fig5 ] , black dots indicate the calibrated ogle - iv stars and green dots indicates the stars in baade s window observed by the _ hst _ @xcite , which are corrected for extinction and reddening with respect to the rcg position in the _ hst _ cmd , @xmath80 @xcite , respectively . we can see that the source is located on the right side of the _ hst _ turn - off stars , which indicates that the source is a sub - giant . assuming the source suffers the same dust extinction and reddening as the rcgs and using the expected extinction - free rcg centroid @xmath81 at this position @xcite , we estimated the extinction - free color and magnitude of the source as @xmath82 . the angular source radius , @xmath83 , is calculated by using the observed @xmath84 and the relation between the limb - darkened stellar angular diameter , @xmath85 , @xmath86 and _ i _ from the results of @xcite , @xmath87 this relation comes from a private communication with boyajian by @xcite . this gives @xmath88 , whose error includes the @xmath89 and the 2% uncertainty in equation 5 . the angular einstein radius and the geometric lens - source relative proper motion , @xmath90 , for close and wide models are derived as follows , @xmath91 @xmath92 @xmath93 @xmath94 although the derived @xmath95 is based on the 3.3 @xmath96 detection of the finite source effect , such a small value indicates that the lens system is likely located in the galactic bulge . the lens physical parameters can not be derived directly because the parallax effect was not detected in this event . we conducted a bayesian analysis to get the probability distribution of the lens properties @xcite , but this analysis implicitly assumes that the planet hosting probability does not depend on the mass of the host star . we used the observed @xmath34 and @xmath33 with the galactic model @xcite . we also used the ogle de - reddened blending flux which is the sum of the lens and unrelated stars derived from the light curve fitting as the upper limit for the lens brightness , @xmath97 and @xmath98 . since we can not distinguish the close and wide models , we combined the probability distribution of these models by weighting the probability distribution of the wide model by @xmath99 , where @xmath100 . figure [ fig6 ] and figure [ fig7 ] shows the probability distribution of the lens properties derived from the bayesian analysis . according to the results , the lens host star is an m or k star with a mass of @xmath5 and its distance is @xmath4 away from the earth , which implies that the lens system is likely to be in the galactic bulge . the mass of the planet is @xmath6 and the projected separation is @xmath7 . if we assume a circular and randomly oriented orbit for the planet , the 3-dimensional semi - major axis is expected to be @xmath101 . we reported the discovery and the analysis of the microlensing event ogle-2013-blg-1761 and found it to be consistent with a planetary lens system . there are two degenerate solutions because the part of the planetary anomaly was sparsely sampled . the best models have planetary mass ratios of @xmath102 for the close model ( @xmath2 ) and @xmath103 for the wide model ( @xmath3 ) . we detect the finite source effect . on the other hand , we do not detect a clear parallax signal in the light curve . our bayesian analysis indicates that the lens system consists of a m or k - dwarf orbited by a super - jupiter mass planet . in the core accretion theory , massive jovian planets are rare around low - mass stars because the relative amount of material for forming planet is small @xcite . there are several massive planets around a low - mass star discovered by microlensing @xcite , and such planets are challenging for the core accretion theory . ogle-2013-blg-1761 lb could be another example of them . but we need to be cautious about the result because the prior we used assumes that host stars of all masses were equally likely to host a planet with the measured mass ratio . if we can detect the lens flux in high resolution follow - up observations , we can determine the lens mass @xcite , but we must be careful not to be confused by other stars such as companions to the lens or source or unrelated stars @xcite . the lens detection can be confirmed by measuring the lens - source relative proper motion @xcite . figure [ fig7 ] shows the probability distribution of _ i_- , _ v_- , _ h_- and _ k_-band magnitude of lens star without extinction . the red vertical lines indicates the magnitude of the source star . the _ i_- and _ v_-band magnitude of the source were derived from the light curve fitting and the _ h_- and _ k_-band magnitude of the source were estimated from the stellar color - color relation in @xcite . the lens is predicted to be fainter than the source by 3 mag and 2 mag of the brightness in the _ h_- and _ k_- band , respectively . this indicates that the lens is too faint to be detected by high resolution follow - up observations until the lens and the source are resolved . but if the lens is on the brighter side of the distribution in figure [ fig7 ] , the lens would be easily detected . for example , ogle-2003-blg-235l@xcite and moa-2011-blg-293l@xcite were estimated to be m dwarfs with masses of @xmath104 and @xmath105 respectively via bayesian analysis . but they were revealed to be k or g dwarfs with the mass of @xmath106 and @xmath107 by high resolution follow - up observations by _ hst _ @xcite and keck telescope @xcite . future high resolution imaging by ground large telescopes or _ hst _ may determine the lens mass or set tighter upper limits . the distance to the lens is @xmath4 , which is likely to be in the galactic bulge . @xcite suggested a possibility of the lack of planets in the galactic bulge . ogle-2013-blg-1761 lb can be added to the growing list of planets discovered by microlensing , such as ogle-2015-blg-0051 lb @xcite , ogle-2014-blg-1760 lb @xcite and ogle-2012-blg-0724 lb @xcite which counters this suggestions . this work contributes to sample of planets found by microlensing which is useful for testing planetary formation theories and the planet distribution in the galaxy . + the moa project is supported by jsps kakenhi grant number jp16h06287 , jsps23340064 and jp15h00781 . the ogle project has received funding from the national science centre , poland , grant maestro 2014/14/a / st9/00121 to au . ogle team thanks profs . m. kubiak and g. pietrzyski , former members of the ogle team , for their contribution to the collection of the ogle photometric data over the past years . ts acknowledges the financial support from the jsps , jsps23103002 , jsps24253004 and jsps26247023 . dpb , ab , and ds acknowledge support from nasa grants nnx13af64 g and nnx16an69 g . work by iab was supported by the marsden fund of the royal society of new zealand , contract no . mau1104 . njr is a royal society of new zealand rutherford discovery fellow . nk is supported by grant - in - aid for jsps fellows . dummy alcock , c. , allsman , r. , alves , d. , et al . 1995 , apj , 454l , 125 batista , v. , gould , a. , dieters , s. , et al . 2011 , a&a , 529 , a102 batista , v. , beaulieu , j .- p . , gould , a. , et al . 2014 , apj , 780 , 54 batista , v. , beaulieu , j .- p . , bennett , d. p. , et al . 2015 , apj , 808 , 170 beaulieu , j .- p . , bennett , d. p. , fouqu , p. , et al . 2006 , natur , 439 , 437 bennett , d. p. , & rhie , s. h. 1996 , apj , 472 , 660 bennett , d. p. , anderson , j. , bond , i. a. 2006 , apj , 647 , l171 bennett , d. p. , bond , i. a. , udalski , a. , et al . 2008 , apj , 684 , 663 bennett , d. p. 2010 , apj , 716 , 1408 bennett , d. p. , bhattacharya , a. , anderson , j. , et al . 2015 , apj , 808 , 169 bennett , d. p. , & rhie , s. h. 1996 , apj , 472 , 660 bensy , t. , yee , j. c. , felzing , s. , et al . 2013 , a&a , 549 , a147 bhattacharya , a. , bennett , d. p. , bond , i. a. , et al . 2016 , aj , 152 , 140 bhattacharya , a. , bennett , d. p. , j. anderson , et al . 2017 , arxiv:1703.06947 borucki , w. j. , koch , d. g. , basri , g. , et al . 2011 , apj , 736 , 19 bond , i. a. , abe , f. , dodd , r. j. , et al . 2001 , mnras , 327 , 868 bond , i. a. , udalski , a. , jaroszyski , m. , et al . 2004 , apjl , 606 , l155 boyajian , t. s. , van belle , g. , & von braun , k. 2014 , aj , 147 , 47 butler , r. p. , wright , j. t. , marcy , g. w. , et al . 2006 , apj , 646 , 505 cassan , a. , kubas , d. , beaulieu , j .- , et al . 2012 , natur , 481 , 167 claret , a. 2000 , a&a , 363 , 1081 dong , s. , gould , a. , udalski , a. , et al . 2009 , apj , 695 , 970 dominik , m. 1999 , a&a , 349,108 fukui , a. , gould , a. , sumi , t. , et al . 2015 , apj , 809 , 74 gonzlez hernndez , j. i. , & bonifacio , p. , et al . 2009 , a&a , 497 , 497 gould , a. 1992 , apj , 392 , 442 gould , a. , & loeb , a. 1992 , apj , 396 , 104 gould , a. 2000 , apj , 5442 , 785 gould , a. , udalski , a. , an , d. , 2006 , apj , 644 , l37 gould , a. , dong , s. , gaudi , b. s. , et al . 2010 , apj , 720 , 1073 han , c. , & gould , a. 1995 , apj , 449 , 521 han , c. , udalski , a. , gould , a. , et al . 2016 , aj , 152 , 95h hirao , y. , udalski , a. , sumi , t. , et al . 2016 , apj , 824 , 139h holtzman , j. a. , watson , a. m. , baum , w. a. , et al . 1998 , aj , 115 , 1946 ida , s. , & lin , d. n. c. 2005 , apj , 625 , 1045 kenyon , s. j. , & hartmann , l. , 1995 , apjss , 101 , 117 koshimoto , n. , udalski , a. , sumi , t. , et al . 2014 , apj , 788 , 128 koshimoto , n. , bennett , d. p. , et al . 2017 , in preparation laughlin , g. , bodenheimer , p. , & adams , f. c. , 2004 , apj , 612 , l73 nataf , d. m. , gould , a. , fouqu , p. , et al . 2013 , apj , 769 , 88 penny , m. , henderson , c. b. , clanton , c. , 2016 , apj , 830 , 150 sako , t. , sekiguchi , t. , sasaki , m. , et al . 2008 , exa , 22 , 51 shin , i. -g . , ryu , y. -h . , udalski , a. , et al . 2016 , jkas , 49 , 73 shvartzvald , y. , maoz , d. , kaspi , s. , et al . 2014 , mnras , 439 , 604 shvartzvald , y. , maoz , d. , udalski , a. , et al . 2016 , mnras , 457 , 4089 spergel , d. , gehrels , n. , baltay , d. , et al . 2015 , arxiv:1503.03757 street , r. a. , choi , j. -y . , tsapras , y. , et al . 2013 , apj , 763 , 67 street , r. a. , udalski , a. , calchi , s. , et al . 2016 , apj , 819 , 93 sumi , t. , abe , f. , bond , i. a. , et al . 2003 , apj , 591 , 204 sumi , t. , bennett , d. p. , bond , i. a. , et al . 2010 , apj , 710 , 1641 sumi , t. , kamiya , k. , bennett , d. p. , et al . 2011 , natur , 473 , 349 suzuki , d. , bennett , d. p. , bond , sumi , t. , et al . 2016 , apj , 833 , 145 tsapras , y. , choi , j. -y . , street , r. a. , et al . 2014 , apj , 782 , 48 udalski , a. , szymaski , m. k. , & szymaski , g. 2015 , acta astron . , 65 , 1 verde , l. , peiris , h. v. , & spergel , d. n. 2003 , apjs , 148 , 195 yee , j. c. , shvartzvald , y. , gal - yam , a. , et al . 2012 , apj , 775 , 102	we report the discovery and the analysis of the planetary microlensing event , ogle-2013-blg-1761 . there are some degenerate solutions in this event because the planetary anomaly is only sparsely sampled . but the detailed light curve analysis ruled out all stellar binary models and shows that the lens to be a planetary system . there is the so - called close / wide degeneracy in the solutions with the planet / host mass ratio of @xmath0 and @xmath1 with the projected separation in einstein radius units of @xmath2 ( close ) and @xmath3 ( wide ) , respectively . the microlens parallax effect is not detected but the finite source effect is detected . our bayesian analysis indicates that the lens system is located at @xmath4 away from us and the host star is an m / k - dwarf with the mass of @xmath5 orbited by a super - jupiter mass planet with the mass of @xmath6 at the projected separation of @xmath7 . the preference of the large lens distance in the bayesian analysis is due to the relatively large observed source star radius . the distance and other physical parameters can be constrained by the future high resolution imaging by ground large telescopes or _ hst_. if the estimated lens distance is correct , this planet provides another sample for testing the claimed deficit of planets in the galactic bulge .
it is generally accepted that interstellar turbulence plays a crucial role in many astrophysical processes , including molecular cloud and star formation , mass and energy transfer in accretion disks , acceleration of cosmic rays , etc . at present we are still groping for the basic facts related to this complex phenomenon and one of the reasons for such an unsatisfactory state of affairs is that the interstellar turbulence statistics are not directly available . for instance , in studies of the neutral medium , indirect measures , such as spectral line widths and centroids of velocity , are employed ( e.g. miesch & bally 1994 ; see also the review by scalo 1987 ) , while potentially more valuable sources of information available in velocity - channel maps remain mostly untapped ( although see heyer & schloerb 1997 ; rosolowsky et al . 1999 ; brunt & heyer 2000 ) because of the difficulty of relating the two - dimensional ( 2d ) statistics available through observations to the underlying three - dimensional ( 3d ) statistics . a discussion of various approaches to the problem of turbulence study using spectral line data can be found in a recent review by lazarian ( 1999 ) . in particular , the problem has been recently addressed by lazarian & pogosyan ( 2000 , herafter lp00 ) , who derived the index ( i.e. , the logarithmic slope ) of the power spectrum - dimensional field does _ not _ include the @xmath0 factor corresponding to a summation over a shell of wavenumbers of radius @xmath1 . with this notation , the well - known kolmogorov @xmath2 law corresponds to an index of @xmath3 . ] of the intensity in velocity channel maps as a function of the corresponding indices for the 3d density and velocity fields of the emitting medium . their work effectively provides a means of inverting the problem , so that the power spectra of the medium s density and velocity can be retrieved from the power spectrum of the emissivity . note that this procedure involves considering channels of various velocity widths or `` velocity slices of different thickness '' , in the terminology of lp00 . otherwise there is an indetermination due to the fact that for sufficiently shallow density spectra the emissivity in slices is given by a linear combination of the velocity and density indices . recent measurements by stanimirivic ( 2000 ) and stanimirivic & lazarian ( 2000 ) have confirmed a variation of the spectral index of hi intensity maps as the slice width is varied , in accordance with the predictions of lp00 . the approach suggested by lp00 allows , in principle , to use the wealth of existing spectroscopic data for deriving interstellar turbulence statistics , thus possibly permitting new levels of understanding of this phenomenon . however , the derivation of lp00 assumes that the statistics of velocity and density are independent . although those authors showed that for a particular case their results hold even when maximal correlation of velocity and density is assumed , testing the scheme on realistic fields from compressible magneto - hydrodynamic ( mhd ) simulations is essential . in this paper we assess the degree of correlation between the density and velocity fields from mhd , compressible turbulence simulations , and whether the theoretical results from lp00 apply in this case , our goal being to find out to what extent the spectra of velocity and density can be recovered from the observed emissivity statistics . note that if the interdependence of velocity and density is important for the recovery of their spectral indices , the inversion must be modified to account for the velocity - density correlations . numerous measurements of the emissivity within actual data cubes suggest that the spectrum follows a power law ( green 1993 , stanimirovic et al . 1999 ) , as is the case in classical high - reynolds number incompressible turbulence ( e.g. , lesieur 1990 ) , and high - resolution numerical simulations of highly compressible mhd turbulence in less than three dimensions ( e.g. , passot , vzquez - semadeni & pouquet 1995 ; gammie & ostriker 1996 ) . a power - law assumption was also used by lp00 . numerical simulations of the ism in 3d , however , usually do not have a large enough inertial range to produce good power laws and this can complicate our interpretation of the results . to deal with this issue , below we describe a procedure for `` correcting '' the spectral indices of the numerically - simulated density and velocity fields while preserving their underlying spatial correlations . we describe the way how our numerical data were produced , and the correlation between density and line - of - sight velocity in section [ sec : num_exp ] , compare numerical results with the predictions of lp00 in section [ sec : stats ] , and provide a general discussion in section [ sec : disc ] . finally , our conclusions are summarized in section [ sec : sum ] . in order to test whether the results from the analytical study of lp00 carry over to the case when the density and velocity fields are correlated according to self - consistent fluid evolution , in the following sections we explicitly calculate the spectral slopes for velocity channel - map data in position - position - velocity ( ppv ) cubes obtained from a three - dimensional simulation of the ism at intermediate size scales ( 3300 pc ) . the simulation is nearly identical to the one presented by pichardo et al . ( 2000 ) , except at slightly lower resolution . both the simulation and the procedure for obtaining the velocity channel data have been discussed at length in that paper , and we refer the reader to it for details . in this section we just outline the information necessary for the purposes of the present paper . channel maps are essentially maps of column density within a given line - of - sight ( los ) velocity interval . throughout this paper we refer to such column density as `` emissivity '' or `` map intensity '' , in analogy to the observational situation . for the sake of simplicity , thermal broadening , whose effects on the statistical analysis are discussed by lp00 , is not considered . the simulation represents a cubic box of 300 pc on a side on the galactic plane , centered at the solar galactocentric distance . the magneto - hydrodynamic equations are solved on a @xmath4 cartesian grid with the @xmath5 , @xmath6 and @xmath7 directions respectively representing the radial , tangential and vertical directions in the galactic disk . a pseudospectral scheme with periodic boundary conditions is used , which includes self - gravity , parameterized heating and cooling , and modeled star formation , such that a `` star '' ( a local heating source which causes its surroundig gas to expand ) is turned on wherever the density exceeds a threshold @xmath8 and @xmath9 . the `` stars '' remain on for 3.7 myr after the criterion is met . this procedure is intended to mimic hii region expansion from ob star ionization heating , and provides an energy source for maintaining the turbulence in the simulation . energy injection by supernovae is not included due to limitations of the numerical scheme , but the total energy injected per stellar event over the stellar lifetime is of the same order of magnitude as that which would be injected by a supernova event . the simulations also include the coriolis force corresponding to a rotation of the frame around the galactic center every @xmath10 yr , and a shear in the @xmath5-@xmath6 plane of the form @xmath11 km s@xmath12 , where @xmath13 is the @xmath5-component of the velocity field . the turbulent motions occur on top of this shearing velocity . this sinusoidal shear is not highly realistic , but is the only one possible with periodic boundary conditions , and was introduced by passot et al . ( 1995 ) as a crude approximation of the galactic shear . due to the periodic boundary conditions , no stratification is present in the @xmath7-direction , but this is not too strong a concern given the range of scales represented by the simulation . because the star formation prescription prevents the density from reaching values significantly larger than @xmath8 , we turn it off after the turbulence is fully developed , and focus on a snapshot of the simulation shortly after that time . at the time of the snapshot , the maximum and minimum values of the ( number ) density are 109 and 0.43 @xmath14 , respectively . a uniform magnetic field of 1.5 @xmath15 g parallel to the @xmath5-direction is included , on top of which turbulent magnetic fluctuations induced by the stellar energy injection occur . the maximum and minimum values of the magnetic field strength are 12.5 and @xmath16 g . an image of the density field is shown in fig . [ fig : den_uz]a . we adopt the @xmath7-direction in the simulation ( perpendicular to the galactic plane ) as the los direction , in order to prevent the shear from introducing power unrelated to the turbulent fluctuations into the velocity spectrum . other than that , since the simulation has no vertical stratification , the @xmath7 direction is statistically equivalent to the @xmath6 direction . only the @xmath5-direction , parallel to the mean magnetic field , is special . figure [ fig : den_uz]b shows an image of the los - component of the velocity field . the structures in the simulations have been discussed at length by pichardo et al.(2000 ) . however , neither the density nor the los - velocity fields have spectra well suited for studying the emissivity spectrum since , due to the low resolution of the simulation , the spectra are not clear power laws , while most turbulence theories , including lp00 s , consider power - law spectra . to circumvent this problem , in what follows we use modified density and velocity fields , such that their spectra are indeed power laws , but the phase coherence of the original fields is preserved . a partial justification for this procedure stems from the fact that the spatial information is contained in the phases of the fourier decomposition of a given field , while the spectrum is related exclusively to the relative amplitudes of the various modes ( armi & flament 1985 ) . unfortunately , this justification is not complete , because the velocity - density coherence does change to some extent as the spectral slope is modified . we discuss this issue in some more detail in [ sec : correl ] . as an indication of what would be realistic indices for the fields , we note that incompressible mhd simulations that resolve the inertial range ( cho & vishniac 2000a , b ) tend to result in a kolmogorov - type spectrum with slope @xmath3 , as theoretically predicted by goldreich & shridhar ( 1995 ) , while highly compressible simulations in less than 3d ( e.g. , passot et al . 1995 ; scalo et al . 1998 [ 2d ] ; gammie & ostriker 1996 [ 1 + 2/3 d ] ) tend to give shock - spectra for the velocity with slopes near @xmath17 and density spectra with slopes near @xmath18 . thus , one may expect spectral indices close to these values also in 3d highly compressible turbulence , so that the spatial correlations from our simulations should probably be most appropriate for those ranges of values . on the other hand , we do not know beforehand what sort of phase coherence should be present for either shallower or steeper spectra , and therefore in these cases our present study is limited to testing whether the formulae by lp00 are correct in the presence of the particular sort of correlations produced by our simulations . the spectral index modification procedure is as follows . we perform a fourier decomposition of the density and los velocity as @xmath19 where is the position vector , @xmath20 and @xmath21 are the fourier amplitudes of the fields ( we drop the subindex @xmath7 in the latter for convenience ) and is the ( vector ) wavenumber . in general , the amplitudes are of the form @xmath22 , where @xmath23 and @xmath24 are respectively the magnitude and phase of @xmath25 . thus , we replace the magnitudes of both fields by new ones ( labeled by primes ) satisfying @xmath26 , where @xmath27 is the desired power - law index . the phases are unchanged . the modified amplitudes are then transformed back to configuration space , rendering new fields which preserve the spatial structure of the original fields but with the desired power - law spectrum . we avoid aliasing by including only modes with wavenumbers lying within a circle of radius equal to half the resolution in wavenumber space in the inverse transform . as an example , in fig . [ fig : old_new_spec ] we show the original power spectrum of the density field together with the resulting modified spectrum for @xmath28 , while fig . [ fig : old_new_dens ] shows cuts through the original and modified density fields , to appreciate the kind of structural changes arising from the spectrum - modification procedure . in the remainder of this paper , we reserve the symbol @xmath27 for the index of the density power spectrum , and denote the velocity power ( or energy ) spectrum index by @xmath15 . using this procedure , we construct a suite of density fields with spectral indices @xmath29 , @xmath17 , and @xmath30 , and of velocity fields with spectral indices @xmath31 , @xmath32 ( the kolmogorov value ) , and @xmath33 . from these , we then construct the corresponding ppv cubes for each density - velocity pair , which are used in the next section to obtain the projected intensity and velocity - channel maps . to characterize the correlation between velocity and density one can use the following function @xmath34 where is the spatial separation ( or `` lag '' ) , is the spatial position , @xmath35 is the density field , is the total three - dimensional velocity vector , and @xmath36 and @xmath37 are respectively the standard deviations of the density and of the velocity . the averaging is performed over all -space and over all angles of the lag vector . this function is similar to the function @xmath38 introduced in lp00 to calculate the effect of maximal allowable velocity - density correlations . the difference between eq.([eq : lp00_corr ] ) and @xmath38 amounts to the normalization and the @xmath39 dependence , where @xmath40 is the angle between @xmath41 and the @xmath7-axis . however , since the projection along precludes computing this correlation using fourier transform techniques , and computing the full correlation numerically in physical space would take prohibitely long times , in this paper we compute a related correlation involving only the los components of both the lags and of the velocity vector . this correlation , given by @xmath42 is much more economical numerically , but contains the same essential information as eq . ( [ eq : lp00_corr ] ) . in eq . ( [ eq : cross_corr ] ) , @xmath43 is the separation along the los and the averaging is performed over all positions @xmath44 in the simulation box and over separations @xmath7 and @xmath45 . this correlation is shown in fig . [ fig : correl]a for the original density and los - velocity fields . as mentioned in [ sec : sims ] , the correlations do change to some extent upon the change in spectral indices . we have observed that the correlation increases as the slopes of either the density or the velocity spectra are made steeper . since a steeper ( resp . shallower ) slope means that the small scales are depressed ( resp . enhanced ) with respect to the large scales , the observation implies that the density and velocity fields are more similar to each other at large scales than at small ones . the correlation between the modified fields most resembles that between the original fields when the modified slopes are closest to those of the original spectra . the correlations observed are moderate , having a maximum @xmath46 ( in absolute value ) for the original fields , and @xmath47 for the modified fields with steepest slopes . in retrospect , this result can be understood in terms of the fact that the fluid equations link one field with the spatial derivatives of the other : in the continuity equation , the density evolution is determined by the _ divergence _ of the velocity field , and in the momentum equation , the velocity evolution is linked to the pressure _ gradient _ , which at best can be related to the density gradient , if the flow behaves in an approximately barotropic way ( e.g. , vzquez - semadeni , passot & pouquet 1996 ) . thus , stronger correlations may be expected between the density and the velocity divergence , and between the velocity and the density gradient , but not so much between the plain density and los - velocity fields . indeed , fig . [ fig : correl]b shows the correlation between the original density field and the divergence of the original 3d velocity field . a larger ( negative ) correlation , of up to @xmath48 is seen , supporting this interpretation . whether or not a density - velocity correlation of the observed strength can alter the results of lp00 is not clear _ a priori _ , and this is the motivation for our study below . unlike the observational situation , in numerical simulations the density and velocity fields are available directly . this allows us to simulate observations and to control the accuracy with which the individual 3d statistics are recovered . spectroscopic observations usually deal with velocity channel maps in which the velocity resolution is determined by the instrument . in the case of our numerical simulation , velocity resolution is not an issue , and we can produce as many velocity channels as desired from the original density and los - velocity data cubes . surely the statistics within an individual slice are degraded as the slice gets thinner , producing progressively higher levels of shot noise , but we compensate for this effect by averaging over the whole set of slices available . to compare our results with the theoretical predictions of lp00 we recall that two possible regimes were found there , depending on the width of the velocity channels . they were termed `` thin '' and `` thick '' slicings , and their emergence is well motivated physically . if we consider turbulence at a scale @xmath49 , the expected squared velocity difference ( the second - order structure function ) between points separated a distance @xmath49 scales as @xmath50 , where @xmath51 is @xmath52 for kolmogorov - type turbulence , and @xmath53 is a constant that depends on the intensity of turbulence ( see , e.g. , lesieur 1990 ) . the structure function index @xmath51 is related to the velocity power spectrum s index @xmath15 by @xmath54 . thus , when this velocity difference is larger than the width of the velocity slice , the slice is considered _ thin _ , and _ thick _ otherwise . lp00 showed that different slopes for the emissivity spectrum are expected in the two regimes mentioned above . it is obvious that whether the slice is either thick or thin with respect to the characteristic turbulent velocity difference depends on the scale being considered , and therefore , since the thin and thick asymptotics differ in general , we expect to see a change of the slope of the emissivity spectrum at some transition scale which depends on the channel width . to test the predictions of lp00 , in fig . [ fig : xi ] we plot the spectra of the emissivity fluctuations in velocity slices of the ppv data cubes ( `` channel maps '' ) ( _ dash - dotted lines _ ) , the spectra of the 3d density fields ( _ solid lines _ ) , and the spectra of the density in thin spatial slices of the 3d density field ( _ dashed lines _ ) , for various combinations of the density and velocity spectral slopes . these plots confirm the theoretical expectations . first of all , the transition from thick to thin slices depends on the spectral index of the velocity . for relatively small numbers of slices we have observed that the slices are still thin for shallow indices and show signatures of becoming effectively thicker for steep indices . the width of the slice is given by @xmath55 , where @xmath56 and @xmath57 is the number of slices . the value of @xmath58 along the los will be a factor of a few higher than the dispersion @xmath59 along the same los . conservatively , we shall assume this factor to be 5 , which corresponds to @xmath60 positive and negative excursions . the squared dispersion over the box size @xmath61 , @xmath62 , equals @xmath63 for power - law statistics . therefore , in order for the slice to be thin , it is required that @xmath64 , where @xmath65 is the scale under consideration in the plane of the slice , and is essentially the inverse of wavenumber in the emissivity spectrum . this means that @xmath66 . another expected feature is the flattening of the spectrum at large @xmath1 ( small separations @xmath65 ) due to shot noise . in our case we derive the slope considering scales from @xmath67 down to @xmath68 , where @xmath61 is the length of the integration box . figure [ fig : xi ] shows that for @xmath65 smaller than @xmath69 the emissivity spectrum indeed flattens . the considerations above mean that for the steepest velocity index of @xmath33 ( i.e. @xmath70 ) , the number of slices should be larger than @xmath71 in order for them to be thin . on the other hand , an increase in the number of slices results in higher shot noise due to the discrete sampling of the velocity along the line of sight ( we only have 100 samples along each line of sight ) . thus , we have adopted a slice width of @xmath72 , which we have found to correspond to thin slicing for the entire parameter space that we explore , but does not introduce too high a level of shot noise . for thin slices and sufficiently _ steep _ density spectra ( @xmath73 ) , lp00 predicted that the measured emissivity spectrum _ does not _ depend on the density spectral slope , and has an index @xmath74 for example , for kolmogorov turbulence , @xmath75 , and therefore an emissivity index of @xmath76 is expected . conversely , for _ shallow _ density spectra , i.e. when @xmath77 , the spectral index in thin slices is expected to be ( lp00 ) @xmath78 table 1 provides a quantitative comparison between the theoretical predictions of lp00 and our numerical results , showing the spectral indices of the emissivity in the velocity channels as a function of the indices of the underlying density and velocity fields . the theoretical predictions of lp00 ( given by equations ( [ steep ] ) and ( [ shallow ] ) ) are indicated in parentheses , while the measured values in the simulated fields are shown in bold type . it is seen that in all cases the agreement is within 10% , for both `` shallow '' and `` steep '' density spectral indices . we stress that the spectra of the 3d density and velocity data cubes , and of the 2d velocity slices , are obtained by integrating with the appropriate configuration space factors , which are @xmath79 and @xmath80 respectively . the main prediction of lp00 is the importance of the velocity statistics in the determination of the emissivity statistics , specifically , its power spectrum . our numerical results confirm this conclusion . indeed , table 1 shows that as the velocity spectral index is varied ( downwards along the columns in table 1 ) , significant variations occur in the emissivity spectral index . moreover , for steep density spectra , the emissivity statistics are roughly independent of the density spectral index . for example , consider the cases with velocity indices @xmath31 and @xmath3 . it is seen that as the underlying density spectral index is changed from @xmath17 to @xmath30 , the emissivity spectral index in thin channels remains constant within the precision of our measurements . for the case with velocity index of @xmath33 , the emissivity index changes slightly ( from @xmath81 to @xmath82 ) for the same variation of the density index as above , but such slight change is within the uncertainty of 0.2 , and brackets the value predicted by the lp00 theory . _ by itself , this result sends a message of warning against attempts to interpret features in the velocity channel maps as actual density enhancements . _ @xmath83 \hline\\ { \rm velocity~index}\\ \hfil\downarrow\hfil\\ \hline\hline\\ -3.2 & { \bf -2.2}~(-2.40 ) & { \bf -2.9}~(-2.90 ) & { \bf -2.9}~(-2.90)\\[1 mm ] \hline \\[1 mm ] -11/3 & { \bf -1.9}~(-2.17 ) & { \bf -2.6}~(-2.67 ) & { \bf -2.6}~(-2.67)\\[1 mm ] \hline\\ -4.5 & { \bf -1.8}~(-1.75 ) & { \bf -2.1}~(-2.25 ) & { \bf -2.3}(-2.25)\\[1 mm ] \hline \end{array}\ ] ] the above discussion suggests that , whatever the velocity spectrum , it can be restored from observations . a possible ambiguity that arises in the case of shallow density spectra , when both the velocity and density contribute to the channel map spectrum , can be resolved by considering several slice widths . for example , it is obvious that if integration is performed over the whole velocity range , the resulting statistics can depend on the density only . the corresponding density inversion is discussed by lazarian ( 1995 ) . for power - law statistics , the spectral index for _ very thick _ velocity - integrated slices equals the spectral index of the underlying 3d density . we have verified this result by integrating over the velocity ( the plots corresponding to the integrated 2d statistics coincide with those of the 3d spectra in fig . [ fig : xi ] ) . we have also confirmed other predictions from lp00 , such as the change of the density spectral index from @xmath27 in volume to @xmath84 in a thin _ spatial _ density slice , as shown in fig . [ fig : xi ] by the dashed lines . another important prediction of lp00 was the gradual steepening of the emissivity spectrum in the velocity slices as their width increases . this prediction was confirmed by stanimirovic ( 2000 ) and stanimirivic & lazarian ( 2000 ) on the basis of 21 cm small magellanic cloud data and fig . [ fig : slice_var ] shows that this indeed happens in our simulation data . in particular , the maximum possible thickening occurs when the velocity information is averaged out . in our case this corresponds to integration through the velocity box , or , equivalently , from an integration throughout the full line of sight in the original 3d density data . it is evident that in this case the velocity fluctuations should not matter and only density statistics should influence the results . we have verified that both a pure density integration along the los and a velocity integration of the emissivity data cubes give the same spectral index . in [ sec : correl ] we have shown that the density and los - velocity fields in our ism simulation are moderately correlated . an important question is whether there are interstellar situations in which the velocity - density correlations are much stronger than in our simulations and whether the results by lp00 are applicable to those situations . the answer to this question may be obtained by analyzing simulations of regions with physical conditions different from those assumed here . however , even in the case of strong correlations , the analytical results of lp00 may hold , as it was shown by lp00 for a particular case of maximal possible velocity - density correlations . further research should clarify the issue . astrophysical implications of the asymptotic relations obtained in lp00 are discussed at length in lazarian ( 1999 ) . here we will stress two points . first , the agreement between our measured emissivity spectral indices and the predictions of lp00 makes us more optimistic about the application of the lp00 technique to emission - line studies of actual interstellar turbulence . second , it also gives us confidence that the interpretation by lp00 of the hi intensity spectrum ( green 1993 ; stanimirovic et al . 1999 ) as arising from velocity fluctuations with a kolmogorov ( i.e. , @xmath3 spectral index ) spectrum is correct . it is worth pointing out that , even though the notion of kolmogorov turbulence has been mentioned frequently throughout the paper , we do not wish to imply that the ism in general exhibits kolmogorov turbulence . arguments why the kolmogorov ( 1941 ) description is not likely to be adequate for the ism have been given by several authors ( e.g. , scalo 1987 ; passot , pouquet & woodward 1988 ; lazarian 1995 ; vzquez - semadeni 1999 ; vzquez - semadeni et al . 2000 ) . briefly speaking , the ism flow , unlike that in the kolmogorov description , is magnetized and compressible . one would think that this should change the scaling . indeed , strong compressibility may give rise to a spectrum of shocks , with a slope of @xmath17 ( in the notation convention used in this paper ) . on the other hand , a theory of magnetized , incompressible turbulence recently put forward by goldreich and sridhar ( 1995 ) predicts the same @xmath85 kolmogorov scaling for motions perpendicular to magnetic field lines . such motions would dominate small - scale asymptotics and therefore the measured spectrum may stay kolmogorov - like , although the nature of the cascade is influenced by magnetic field . moreover , the interpretation of the observational data given in lp00 , and supported in this paper , is suggestive that the kolmogorov scaling may ultimately survive compressibility or , alternatively , that the hi data considered there sample gas that is only weakly compressible . further work , both theoretical and observational , is needed to resolve this issue . in particular , application of the techniques used here to molecular - line data should prove of great interest . to conclude we should stress that for kolmogorov turbulence there exists a _ coincidence _ between the spectral index of a thin , 2d spatial slice of the density , which is : @xmath86 and the spectral index of the emissivity fluctuations in a thin velocity channel when the density spectrum is steep ( e.g. kolmogorov ) and the velocity spectrum follows a kolmogorov law . indeed , the former , calculated using eq . ( [ 2ddensity ] ) , is @xmath87 while the latter , calculated using eq . ( [ steep ] ) , takes the same value @xmath88 . although the accidental character of this coincidence was mentioned in lp00 , our experience shows that it does cause confusion . for instance , we have had to confront the point of view that the observed spectrum of hi fluctuations is a simple consequence of eq . ( [ 2ddensity ] ) . this is a _ fallacy _ and , while we do not wish to repeat the arguments of lp00 , table 1 shows that the naive `` rule '' given by eq . ( [ 2ddensity ] ) does not return the correct emissivity spectral indices for velocity spectra other than kolmogorov s , and that the coincidence is accidental for the kolmogorov index . the correct formula for calculating the spectral index of the emissivity fluctuations in thin slices when the density spectrum is steep is given by eq . ( [ steep ] ) . in this paper we have used data from a numerical simulation of compressible mhd turbulence in the ism to study the correlation between the density and los - velocity field , and the dependence of the spectral index of the velocity - channel emissivity on the indices of the original three - dimensional velocity and density fields . to do so , we have _ a ) _ computed the cross - correlation between the density and los - velocity fields , _ b ) _ modified their spectral slopes to pre - determined values , and _ c ) _ produced ppv data from them . from the ppv data , we computed in turn the emissivity spectrum in velocity slices of those cubes ( `` channel maps '' ) , allowing us to directly test the results from the analytical study by lp00 . we have found that its predictions hold also for the case of fields obeying the fluid conservation equations . in particular , we have found that for steep density spectra with power - law indices @xmath73 , the emissivity spectral index does not depend on the actual value of the density index and is determined exclusively by the velocity fluctuations . this indicates that only the velocity field is responsible for the structure existing in thin slices of the ppv cubes if the density spectrum is steep enough , in agreement with the previous result by pichardo et al . ( 2000 ) that the morphology in the channel maps is more resemblant of that of the los - velocity field than that of the density field , and with previous suggestions that `` objects '' identified in position - velocity maps may not correspond to actual density features ( adler & roberts 1992 ; pichardo et al . 2000 ; ostriker , stone & gammie 2000 ) . the spectral index measured for the emissivity fluctuations in the numerical simulation was within 10% of the value given by relation ( [ steep ] ) , derived by lp00 , in which @xmath51 is the exponent of the second - order structure function . for a shallow density spectrum , namely @xmath77 , we found that the emissivity index depends on both velocity and density . again the lp00 expression for the resulting power - law index , relation ( [ shallow ] ) , provides a satisfactory description of our measurements . we also estimated the accuracy of our measurements both by integrating out the velocity within ppv data cubes and by taking thin slices of the density data cubes . these cases have straightforward analytical solutions ( see lp00 ) and served as benchmarks . we also verified that variations of the emissivity spectrum resulting from the change of the velocity slice width ( see fig.[fig : slice_var ] ) are very similar to those reported in the studies by stanimirovic ( 2000 ) and stanimirovic & lazarian ( 2000 ) , in which the width of observational hi data slices was varied to test the predictions from lp00 . in brief , our results imply that : + 1 . velocity creates small scale structure within slices of ppv data cubes and therefore the interpretation of features seen in ppv data cubes as density structures ( `` clouds '' ) may be misleading . + 2 . for sufficiently shallow density spectra , velocity dominates the statistics of intensity fluctuations within ppv slices . emissivity spectra become steeper as the width of the velocity slices increases and , for sufficiently thick slices , the spectra are dominated by density enhancements only . part of this work was completed while two of us ( a.l . and e.v .- s . ) participated in the astrophysical turbulence program of the institute for theoretical physics at the university of california at santa barbara . the numerical simulation was performed on the cray y - mp 4/64 of dgsca , unam . this work has received partial funding from conacyt , mxico , through grant 27752-e to e.v .- s . and from nsf , usa , through grant phy94 - 07194 . adler , d. d. , & roberts , w. w. 1992 , apj , 384 , 95 armi , l. & flament , p. 1985 , j. geophys . research , 90 , no . c6 , 11779 brunt , c. & heyer , m. h. 2000 , preprint ( astro - ph/0011200 ) cho & vishniac 2000a , apj , 539 , 253 cho & vishniac 2000b , apj , 538 , 217 gammie , c.f . and ostriker , e.c . 1996 , apj 466 , 814 green , d.a . 1993 mnras , 262 , 328 goldreich , p. & sridhar , s. 1995 , apj , 438 , 763 kolmogorov , a. 1941 , compt . ussr , 30 , 301 heyer , m. h. & schloerb . f.p 1997 , apj , 475 , 173 lazarian , a. 1995 , a&a , 293 , 507 lazarian , a. 1999 , in plasma turbulence and energetic particles , ed . m. ostrowski and r. schlickeiser , cracow , 28 , astro - ph/0001001 lazarian & pogosyan 2000 , apj , 537 , 720 lesieur , m. 1990 , turbulence in fluids , 2nd ed.(dordrecht : kluwer ) miesch , m.s . , & bally , j. 1994 , apj , 429 , 645 ostriker , e.c . , stone , j.m . & gammie , c.f . 2000 ( preprint : astro - ph/0008454 ) passot , t. , pouquet , a. & woodward , p. 1988 , a&a passot , t. , vzquez - semadeni , e. , & pouquet , a. 1995 pichardo , b. , vzquez - semadeni , e. , gazol , a. , passot , t. & ballesteros - paredes , j. 2000 , apj , 532 , 353 rosolowsky e.w . , goodman , a.a . , wilner , d.j . , & williams , j.p . 1999 , apj , 524 , 887 scalo , j.m . 1987 , in interstellar processes , eds . hollenbach & h.a . thronson , reidel , dordreicht , 349 scalo , j.m . , vzquez - semadeni , e. , chappell , d. & passot , t. 1998 , apj , 504 , 835 stanimirovic , s. 2000 , in the evolution of galaxies : observational clues " , eds . vilchez , g. stasinska & e. perez ( in press ) stanimirovic , s. & lazarian , a. 2000 , apj , submitted stanimirovic , s. , staveley - smith , l. , dickey , j.m . , sault , r.j . , & snowden , s.l . 1999 , mnras , 302 , 417 vzquez - semadeni , e. , passot , t. & pouquet , a. 1996 , apj , 473 , 881 vzquez - semadeni , e. 1999 , in millimeter & submillimeter astronomy : chemistry and physics in molecular clouds , eds . w. f. wall , a. carramiana , l. carrasco , and p. f. goldsmith ( dordrecht : kluwer ) , 161 vzquez - semadeni , e. , ostriker , e. c. , passot , t. , gammie , c. & stone , j. 2000 , in `` protostars & planets iv '' , eds . v. mannings , a. boss & s. russell ( tucson : univ . of arizona press ) , 3	in this paper we test the results of a recent analytical study by lazarian and pogosyan , on the statistics of emissivity in velocity channel maps , in the case of realistic density and velocity fields obtained from numerical simulations of mhd turbulence in the interstellar medium ( ism ) . to compensate for the lack of well - developed inertial ranges in the simulations due to the limited resolution , we apply a procedure for modifying the spectral slopes of the fields while preserving the spatial structures . we find that the density and velocity are moderately correlated in space and prove that the analytical results by lazarian and pogosyan hold in the case when these fields obey the fluid conservation equations . our results imply that the spectra of velocity and density can be safely recovered from the position - position - velocity ( ppv ) data cubes available through observations , and confirm that the relative contributions of the velocity and density fluctuations to those of the emissivity depend on the velocity resolution used and on the steepness of the density spectral index . furthermore , this paper supports previous reports that an interpretation of the features in the ppv data cubes as simple density enhancements ( i.e. , `` clouds '' ) can be often erroneous , as we observe that changes in the velocity statistics substantially modify the statistics of emissivity within the velocity data cubes . * u *
this paper deals with combinatorial aspects in the representation theory of algebras . more precisely , for certain classes of algebras which are defined purely combinatorially by directed graphs and homogeneous relations we will characterize important representation - theoretic invariants in a combinatorial way . in particular , this leads to new explicit invariants for the derived module categories of the algebras involved . the starting point for this article is that the unimodular equivalence class of the cartan matrix of a finite dimensional algebra is invariant under derived equivalence . hence , being able to determine normal forms of cartan matrices yields invariants of the derived category . the class of algebras we study are the gentle algebras , and the related skewed - gentle algebras . gentle algebras are defined purely combinatorially in terms of a quiver with relations ( for details , see section [ sec - def ] ) ; the more general skewed - gentle algebras ( introduced in @xcite ) are then defined from gentle algebras by specifying special vertices which are split for the quiver of the skewed - gentle algebra ( see section [ sec - skewed ] ) . these algebras occur naturally in the representation theory of finite dimensional algebras , especially in the context of derived categories . for instance , the algebras which are derived equivalent to hereditary algebras of type @xmath2 are precisely the gentle algebras whose underlying undirected graph is a tree @xcite . the algebras which are derived equivalent to hereditary algebras of type @xmath3 are certain gentle algebras whose underlying graph has exactly one cycle @xcite . remarkably , the class of gentle algebras is closed under derived equivalence @xcite ; but note that the class of skewed - gentle algebras is not closed under derived equivalence . a fundamental distinction in the representation theory of algebras is given by the representation type , which can be either finite , tame or wild . in the modern context of derived categories , also derived representation types have been defined . again , gentle algebras occur naturally in this context . d. vossieck @xcite showed that an algebra @xmath4 has a discrete derived category if and only if either @xmath4 is derived equivalent to a hereditary algebra of type @xmath2 , @xmath5 , @xmath6 or @xmath4 is gentle with underlying quiver @xmath7 having exactly one ( undirected ) cycle and the number of clockwise and of counterclockwise paths of length 2 in the cycle that belong to @xmath8 are different . skewed - gentle algebras are known to be of derived tame representation type ( for a definition of derived tameness , see @xcite ) . it is a long - standing open problem to classify gentle algebras up to derived equivalence . a complete answer has only been obtained for the derived discrete case @xcite . the main problem is to find good invariants of the derived categories . in this paper we provide easy - to - compute invariants of the derived categories of skewed - gentle algebras which are of a purely combinatorial nature . our results are obtained from a detailed computation of the @xmath0-cartan matrices of gentle and skewed - gentle algebras , respectively . the following notion will be crucial throughout the paper . let @xmath7 be a ( gentle ) quiver with relations . an oriented path @xmath9 with arrows @xmath10 in @xmath11 is called an oriented @xmath12-cycle with full zero relations if @xmath13 has the same start and end point , and if @xmath14 for all @xmath15 and also @xmath16 . such a cycle is called minimal if the arrows @xmath17 on @xmath13 are pairwise different . we call two matrices @xmath18 with entries in a polynomial ring @xmath19 $ ] unimodularly equivalent ( over @xmath19 $ ] ) if there exist matrices @xmath20 over @xmath19 $ ] of determinant 1 such that @xmath21 . we can now state our main result on gentle algebras . * theorem 1 . * [ thm1 ] _ let @xmath7 be a gentle quiver , and @xmath22 the corresponding gentle algebra . denote by @xmath23 the number of minimal oriented @xmath12-cycles in @xmath11 with full zero relations . + then the @xmath0-cartan matrix @xmath24 is unimodularly equivalent ( over @xmath25 $ ] ) to a diagonal matrix with entries @xmath26 , with multiplicity @xmath27 , @xmath28 , and all further diagonal entries being @xmath29 . _ this theorem has the following direct consequences . * corollary 1 . * _ let @xmath7 be a gentle quiver , and @xmath30 the corresponding gentle algebra . denote by @xmath23 the number of minimal oriented @xmath12-cycles in @xmath11 with full zero relations . then the @xmath0-cartan matrix @xmath24 has determinant @xmath31 _ the following consequence of corollary 1 was first proved in @xcite . for a gentle quiver @xmath7 we denote by @xmath32 the number of minimal oriented cycles of odd length in @xmath11 having full zero relations , and by @xmath33 the number of analogous cycles of even length . * corollary 2 . * _ let @xmath7 be a gentle quiver , and @xmath30 the corresponding gentle algebra . then for the determinant of the cartan matrix @xmath34 the following holds . @xmath35 _ the most important application of theorem 1 is the following corollary which gives for gentle algebras easy - to - check combinatorial invariants of the derived category . * corollary 3 . * _ let @xmath7 and @xmath36 be gentle quivers , and let @xmath30 and @xmath37 be the corresponding gentle algebras . if @xmath4 and @xmath38 are derived equivalent , then @xmath39 and @xmath40 . _ as an illustration we give in section [ sec : gentles ] a complete derived equivalence classification of gentle algebras with two simple modules and of gentle algebras with three simple modules and cartan determinant 0 . our main result on skewed - gentle algebras determines the normal form of their @xmath0-cartan matrices . * theorem 2 . * _ let @xmath41 be a skewed - gentle algebra , arising from choosing a suitable set of special vertices in the gentle quiver @xmath7 . denote by @xmath23 the number of minimal oriented @xmath12-cycles in @xmath11 with full zero relations . + then the @xmath0-cartan matrix @xmath42 is unimodularly equivalent to a diagonal matrix with entries @xmath43 , with multiplicity @xmath27 , @xmath28 , and all further diagonal entries being @xmath29 . _ as an immediate consequence we obtain that the cartan determinant of any skewed - gentle algebra is the same as the cartan determinant for the underlying gentle algebra . * corollary 4 . * _ let @xmath41 be a skewed - gentle algebra , arising from choosing a suitable set of special vertices in the gentle quiver @xmath7 , with corresponding gentle algebra @xmath30 . then @xmath44 , and thus in particular , the determinants of the ordinary cartan matrices coincide , i.e. , @xmath45 . _ the paper is organized as follows . in section [ sec - def ] we collect the necessary background and definitions about quivers with relations and ( @xmath0-)cartan matrices . in section [ sec : gentles ] we prove all the main results about @xmath0-cartan matrices for gentle algebras . here we also give some extensive examples to illustrate our results . section [ sec - skewed ] contains the analogous main results for skewed - gentle algebras . * acknowledgement . * we thank the referee for very helpful and insightful comments . in particular , we are grateful for pointing out the importance of graded derived equivalences in our context of @xmath0-cartan matrices . algebras can be defined naturally from a combinatorial setting by using directed graphs . a finite directed graph @xmath11 is called a _ quiver_. for any arrow @xmath46 in @xmath11 we denote by @xmath47 its start vertex and by @xmath48 its end vertex . an oriented path @xmath13 in @xmath11 of length @xmath49 is a sequence @xmath50 of arrows @xmath51 such that @xmath52 for all @xmath53 . ( note that for each vertex @xmath54 in @xmath55 we allow a trivial path @xmath56 of length @xmath57 , having @xmath54 as its start and end vertex . ) for such a path @xmath13 we then denote by @xmath58 its start vertex and by @xmath59 its end vertex . the path algebra @xmath60 , where @xmath61 is any field , has as basis the set of all oriented paths in @xmath11 . the multiplication in the algebra @xmath60 is defined by concatenation of paths , i.e. , the product of two paths @xmath13 and @xmath0 is defined to be the concatenated path @xmath62 if @xmath63 , and zero otherwise . more general algebras can be obtained by introducing relations on a path algebra . an ideal @xmath64 is called admissible if @xmath65 where @xmath66 is the ideal of @xmath60 generated by the arrows of @xmath11 . the pair @xmath7 where @xmath11 is a quiver and @xmath64 is an admissible ideal is called a _ quiver with relations_. for any quiver with relations @xmath7 , we can consider the factor algebra @xmath30 , where @xmath61 is any field . we identify paths in the quiver @xmath11 with their cosets in @xmath4 let @xmath67 denote the set of vertices of @xmath11 . for any @xmath68 there is a path @xmath69 of length zero . these are primitive orthogonal idempotents in @xmath4 , the sum @xmath70 is the unit element in @xmath4 . in particular we get @xmath71 , hence the ( right ) @xmath4-modules @xmath72 are the indecomposable projective @xmath4-modules . the _ cartan matrix _ @xmath73 of an algebra @xmath30 is the @xmath74-matrix defined by setting @xmath75 . recall that when @xmath8 is generated by monomials , @xmath30 is called a monomial algebra . for monomial algebras , computing entries of the cartan matrix reduces to counting paths in the quiver @xmath11 which are nonzero in @xmath4 . in fact , any homomorphism @xmath76 of right @xmath4-modules is uniquely determined by @xmath77 , the @xmath61-vector space generated by all paths in @xmath11 from vertex @xmath78 to vertex @xmath79 , which are nonzero in @xmath30 . in particular , we have @xmath80 . this is the key viewpoint in this paper , enabling us to obtain results on the representation - theoretic cartan invariants by combinatorial methods . it allows to study a refined version of the cartan matrix , which we call the @xmath0-cartan matrix . ( it also occurred in the literature as filtered cartan matrix , see for instance @xcite . ) let @xmath11 be a quiver and assume that the relation ideal @xmath8 is generated by homogeneous relations , i.e. , by linear combinations of paths having the same length ( actually , for the algebras considered in this paper , the ideal @xmath8 will always be generated by monomials and commutativity ( mesh ) relations ) . the path algebra @xmath60 is a graded algebra , with grading given by path lengths . since @xmath8 is homogeneous , the factor algebra @xmath30 inherits this grading . so the morphism spaces @xmath81 become graded vector spaces . recall that the dimensions of these vector spaces are the entries of the ( ordinary ) cartan matrix . * definition . * let @xmath30 be a finite - dimensional algebra , and assume that the ideal @xmath8 is generated by homogeneous relations . for any vertices @xmath78 and @xmath79 in @xmath11 let @xmath82 be the graded components . let @xmath0 be an indeterminate . the @xmath0-cartan matrix @xmath83 of @xmath4 is defined as the matrix with entries @xmath84 $ ] . in other words , the entries of the @xmath0-cartan matrix are the poincar polynomials of the graded homomorphism spaces between projective modules . loosely speaking , when counting paths in the quiver of the algebra , each path is weighted by some power of @xmath0 according to its length . clearly , specializing @xmath1 gives back the usual cartan matrix @xmath34 ( i.e. , we forget the grading ) . even if we are mainly interested in the ordinary cartan matrix , the point of view of @xmath0-cartan matrices provides some new insights as we take a closer look at the invariants of the cartan matrix . [ brauer - ex ] _ we consider the following two quivers . _ let @xmath85 , where the ideal @xmath86 is generated by @xmath87 , @xmath88 and @xmath89 . the @xmath0-cartan matrix of @xmath4 has the form @xmath90 the second algebra @xmath91 is defined by the quiver @xmath92 , subject to the generating relations @xmath93 ( i.e. all paths of length four are zero ) . the @xmath0-cartan matrix of @xmath94 has the form @xmath95 cartan matrices provide invariants which are preserved under derived equivalences and thus improve our understanding of derived module categories ; this is our main motivation to study normal forms , invariant factors and determinants of cartan matrices in this paper . the following result is contained in the proof of ( * ? ? ? * proposition 1.5 ) . [ unimod ] let @xmath4 be a finite - dimensional algebra . the unimodular equivalence class of the cartan matrix @xmath34 is invariant under derived equivalence . in particular , the determinant of the cartan matrix is invariant under derived equivalence . we emphasize that the above theorem only deals with ordinary cartan matrices @xmath96 . the determinant of the @xmath0-cartan matrix is in general not invariant under derived equivalence . as an example , consider the algebras @xmath4 and @xmath94 from example [ brauer - ex ] , with @xmath97 and @xmath98 . but , in fact , the algebras @xmath4 and @xmath94 are derived equivalent ; they are brauer tree algebras for trees with the same number of edges and the same exceptional multiplicity @xcite . note that when specializing @xmath1 we indeed get the same determinants for the ordinary cartan matrices , as predicted by theorem [ unimod ] however , the natural setting when dealing with @xmath0-cartan matrices is that of graded derived categories . indeed , the determinant of the @xmath0-cartan matrix ( which is defined so as to take the grading into account ) is invariant under graded derived equivalences . we are very grateful to the referee for pointing this out to us . we do not discuss this aspect in this paper further , but shall address the topic of graded derived equivalences for gentle algebras in detail in a subsequent publication . for instance , the above algebra @xmath94 is graded derived equivalent to the algebra @xmath4 , where the grading on @xmath4 is chosen so that @xmath46 and @xmath99 are of degree 2 , and @xmath100 and @xmath101 of degree 1 . then rickard s derived equivalence @xcite lifts to a graded derived equivalence . in this section , we shall prove theorem [ thm1 ] on the unimodular equivalence class of the @xmath0-cartan matrix of an arbitrary gentle algebra . we first recall the definition of special biserial algebras and of gentle algebras , as these details will be crucial for what follows . let @xmath11 be a quiver and @xmath8 an admissible ideal in the path algebra @xmath60 . we call the pair @xmath7 a special biserial quiver ( with relations ) if it satisfies the following properties . \(i ) each vertex of @xmath11 is starting point of at most two arrows , and end point of at most two arrows . \(ii ) for each arrow @xmath46 in @xmath11 there is at most one arrow @xmath99 such that @xmath102 , and at most one arrow @xmath101 such that @xmath103 . a finite - dimensional algebra @xmath4 is called special biserial if it has a presentation as @xmath30 where @xmath7 is a special biserial quiver . gentle quivers form a subclass of the class of special biserial quivers . a pair ( @xmath104 as above is called a _ gentle quiver _ if it is special biserial and moreover the following holds . \(iii ) the ideal @xmath8 is generated by paths of length 2 . \(iv ) for each arrow @xmath46 in @xmath11 there is at most one arrow @xmath105 with @xmath106 such that @xmath107 , and there is at most one arrow @xmath108 with @xmath109 such that @xmath110 . a finite - dimensional algebra @xmath4 is called gentle if it has a presentation as @xmath30 where @xmath7 is a gentle quiver . the following lemma will turn out to be very useful . it does not only hold for gentle algebras but for those where we have dropped the final condition ( iv ) in the definition of gentle quivers . recall that two matrices @xmath111 , @xmath112 with entries in @xmath19 $ ] are called unimodularly equivalent ( over @xmath19 $ ] ) if there exist matrices @xmath113 , @xmath11 over @xmath19 $ ] of determinant 1 such that @xmath21 . [ lem : arrow - removal ] let @xmath7 be a special biserial quiver , and assume that @xmath8 is generated by paths of length 2 . let @xmath30 be the corresponding special biserial algebra . let @xmath46 be an arrow in @xmath11 , not a loop , such that there is no arrow @xmath114 with @xmath115 and @xmath116 , or there is no arrow @xmath117 with @xmath118 and @xmath119 . let @xmath120 be the quiver obtained from @xmath11 by removing the arrow @xmath46 , let @xmath121 be the corresponding relation ideal and @xmath37 . then the @xmath0-cartan matrices @xmath24 and @xmath122 are unimodularly equivalent ( over @xmath25 $ ] ) . * we consider the case where @xmath123 is an arrow in @xmath11 such that there is no arrow @xmath114 with @xmath115 and @xmath124 ; the second case is dual . + let @xmath125 . as @xmath7 is special biserial , there is a unique maximal non - zero path starting with @xmath126 , say @xmath127 , where @xmath128 , @xmath129 . as @xmath4 is finite - dimensional , the condition on @xmath130 guarantees that @xmath131 for all @xmath132 , but we may have @xmath133 for some @xmath134 . now any non - zero path of length @xmath79 , say , ending at @xmath135 can uniquely be extended to a non - zero path of length @xmath136 ending at @xmath137 , by concatenation with @xmath138 . conversely , any non - zero path ending at @xmath137 and involving @xmath126 arises in this way . + now denote the column corresponding to a vertex @xmath54 in the @xmath0-cartan matrix @xmath24 by @xmath139 . we perform column transformations on @xmath24 by replacing the columns @xmath140 by @xmath141 , for @xmath142 ( if @xmath133 for some @xmath143 , the column @xmath144 will then be replaced by @xmath145 ) . the resulting matrix @xmath146 is then exactly the cartan matrix @xmath122 to the algebra @xmath38 corresponding to the quiver @xmath120 where @xmath130 has been removed . @xmath147 + for any vertex in a quiver @xmath11 , its _ valency _ is defined as the number of arrows attached to it , i.e. , the number of incoming arrows plus the number of outgoing arrows ( note that in particular any loop contributes twice to the valency ) . [ thm : gentle ] let @xmath7 be a gentle quiver , and let @xmath30 be the corresponding gentle algebra . denote by @xmath23 the number of minimal oriented @xmath12-cycles in @xmath11 with full zero relations . + then the @xmath0-cartan matrix @xmath24 is unimodularly equivalent ( over @xmath25 $ ] ) to a diagonal matrix with entries @xmath26 , with multiplicity @xmath27 , @xmath28 , and all further diagonal entries being @xmath29 . * * we want to prove the claim by double induction on the number of vertices and the number of arrows . clearly the result holds if @xmath11 has no arrows or if it consists of one vertex with a loop . if @xmath11 has a vertex @xmath54 of valency 1 or 3 , or of valency 2 but with no zero relation at @xmath54 , then we can use lemma [ lem : arrow - removal ] to remove an arrow from @xmath11 ; note that by the conditions in lemma [ lem : arrow - removal ] the removed arrow is not involved in any oriented cycle with full zero relations . hence @xmath24 is unimodularly equivalent to @xmath122 , where the corresponding quiver has one arrow less but the same number of oriented cycles with full zero relations , and hence the result holds by induction . hence we may now assume that all vertices are of valency 0 , 2 or 4 , and if a vertex is of valency 2 , then there is a zero relation at the vertex . also , if @xmath11 is not connected , we may use induction on the number of vertices to have the result for the components and thus for the whole quiver ; hence we may assume that @xmath11 is connected . in particular , we now only have vertices @xmath54 of valency 2 with a non - loop zero relation at @xmath54 , and vertices of valency 4 . as we do not have paths of arbitrary lengths , not all vertices can be of valency 4 ( see also ( * ? ? ? * lemma 3 ) ) . now we take a vertex @xmath148 of valency 2 , with incoming arrow @xmath149 and outgoing arrow @xmath150 with @xmath151 ( here , @xmath152 ) . + as @xmath7 is gentle , there is a unique maximal path @xmath13 in @xmath11 with non - repeating arrows starting in @xmath135 with @xmath126 , such that the product of any two consecutive arrows is zero in @xmath4 ; in our present situation this path is an oriented cycle @xmath153 with full zero relations returning to @xmath135 . we denote the vertices on this path by @xmath154 , and the arrows by @xmath128 , @xmath155 ( also @xmath156 ) ; note that the arrows on @xmath13 are distinct , but the vertices are not necessarily distinct ( but we point out that @xmath157 for all @xmath158 ) . + denote by @xmath159 the row of the @xmath0-cartan matrix @xmath24 corresponding to the vertex @xmath160 . in @xmath24 , we now replace the row @xmath161 by the linear combination @xmath162 to obtain a new matrix @xmath146 ( note that this is a unimodular transformation over @xmath25 $ ] ) . + the careful choice of the coefficients is just made so that we can refine the argument in @xcite . we recall some of the notation there . for any arrow @xmath123 in @xmath11 let @xmath163 be the set of paths starting with @xmath123 which are non - zero in @xmath4 . at each vertex @xmath137 there is at most one outgoing arrow @xmath164 ; for this arrow we have @xmath165 , as @xmath7 is gentle . hence , cancelling @xmath166 induces a natural bijection @xmath167 , for @xmath168 , such that a path of @xmath0-weight @xmath169 is mapped to a path of @xmath0-weight @xmath170 ( if there is no arrow @xmath171 , we set @xmath172 ) . as @xmath54 is of valency 2 , with a zero - relation at @xmath54 , we also have the trivial bijection @xmath173 , again with a weight reduction by @xmath0 . now almost everything cancels in @xmath174 , apart from the one term @xmath175 that we obtain as the entry in the column corresponding to @xmath54 . + in the next step , we use the dual ( counter - clockwise ) operation on the columns labelled by the vertices on the cycle @xmath176 , i.e. , we set @xmath177 and replace the column @xmath178 by the linear combination @xmath179 ordering vertices so that @xmath54 corresponds to the first row and column of the cartan matrix , we have thus unimodularly transformed @xmath24 to a matrix of the form @xmath180 where @xmath181 is the @xmath0-cartan matrix of the gentle algebra @xmath38 for the quiver @xmath120 obtained from @xmath11 by removing @xmath54 and the arrows incident with @xmath54 . note that in comparison with @xmath11 , the quiver @xmath120 has one vertex less and one cycle with full zero relations of length @xmath182 less ; now by induction , the result holds for @xmath183 , and hence the result for @xmath24 follows immediately . @xmath147 + this result has several immediate nice consequences . [ q - det ] let @xmath7 be a gentle quiver , and let @xmath30 be the corresponding gentle algebra . denote by @xmath23 the number of minimal oriented @xmath12-cycles in @xmath11 with full zero relations . then the @xmath0-cartan matrix @xmath24 has determinant @xmath31 [ orient - cycles ] let @xmath7 be a gentle quiver , with set of vertices @xmath67 . then , as a direct consequence of theorem [ thm : gentle ] , there are at most @xmath184 minimal oriented cycles with full zero relations in the quiver ( this could also be proved directly by induction ) . note that the property of being gentle is invariant under derived equivalence @xcite , and we now have some invariants to distinguish the derived equivalence classes . for a gentle quiver @xmath7 , recall that @xmath33 and @xmath32 denote the number of minimal oriented cycles in @xmath11 with full zero relations of even and odd length , respectively . as an immediate consequence of corollary [ q - det ] we obtain the following formula for the cartan determinant which was the main result in @xcite : [ det ] let @xmath7 be a gentle quiver , and let @xmath30 be the corresponding gentle algebra . then for the determinant of the cartan matrix @xmath34 the following holds . @xmath185 note that in combination with remark [ orient - cycles ] this implies that the cartan determinant of a gentle algebra @xmath30 is at most @xmath186 , where @xmath187 is the number of simple modules of @xmath4 . the most important application of theorem [ thm : gentle ] is the following corollary which gives for gentle algebras new , combinatorial and easy - to - check invariants of the derived category . [ ec - oc ] let @xmath7 and @xmath36 be gentle quivers , and let @xmath30 and @xmath37 be the corresponding gentle algebras . if @xmath4 and @xmath38 are derived equivalent , then @xmath39 and @xmath40 . * since @xmath4 and @xmath38 are derived equivalent , their ( ordinary ) cartan matrices @xmath34 and @xmath188 are unimodularly equivalent over @xmath189 . by specializing to @xmath1 in theorem [ thm : gentle ] , representatives for the equivalence classes are given by diagonal matrices with entries @xmath190 for each minimal oriented cycle with full zero relations of odd length , an entry @xmath191 for each such cycle of even length , and remaining entries @xmath192 . these are precisely the elementary divisors over @xmath189 . the elementary divisors of an integer matrix are uniquely determined , and the diagonal matrices in theorem [ thm : gentle ] are actually the smith normal forms of @xmath34 and @xmath188 over @xmath189 . but by theorem [ unimod ] the unimodular equivalence class , and hence the smith normal form , is invariant under derived equivalence . hence , the diagonal entries in the above normal forms for @xmath34 and @xmath188 must occur with exactly the same multiplicities . thus we get the same number of minimal oriented cycles with full zero relations of even length and of odd length , respectively , i.e. , @xmath39 and @xmath40 . @xmath147 + we now illustrate our results and apply them to derived equivalence classifications of gentle algebras . [ derived - two ] * gentle algebras with two simple modules . * _ there are nine connected gentle quivers @xmath7 with two vertices , as given in the following list . the dotted lines indicate the zero relations generating the admissible ideal @xmath8 . _ in @xcite it was shown that these are precisely the basic connected algebras with two simple modules which are derived tame . as a direct illustration of our results we show how to classify these algebras up to derived equivalence . recall that the property of being gentle is invariant under derived equivalence @xcite . moreover , the number of simple modules of an algebra is a derived invariant @xcite . thus we will be able to describe the complete derived equivalence classes . we have shown above that the numbers @xmath32 and @xmath33 are derived invariants . in addition we look at two classical invariants , the center and the first hochschild cohomology group @xmath193 . recall that the center of an algebra ( and more generally the hochschild cohomology ring ) is invariant under derived equivalence @xcite . if the quiver contains a loop , then the dimension of @xmath193 depends on the characteristic being 2 or not . we indicate the dimension in characteristic 2 in parantheses in the table below . they can be computed using a method based on work of m. bardzell @xcite on minimal projective bimodule resolutions for monomial algebras ; a very nice explicit combinatorial description is given by c. strametz ( * ? ? ? * proposition 2.6 ) . @xmath194 the algebras @xmath195 are pairwise not derived equivalent . this can be deduced directly from the above table , since the dimensions of the first hochschild cohomology groups are different . the algebras @xmath196 and @xmath197 are derived equivalent . ( this can be shown by explicitly constructing a suitable tilting complex , similar to the detailed example given in the appendix . ) the algebras @xmath198 and @xmath199 with cartan determinant 0 are not derived equivalent , since their centers have different dimensions . in summary , there are exactly eight derived equivalence classes of connected gentle algebras with two simple modules . they are indicated by the double vertical lines in the above table . [ derived - three ] * gentle algebras with three simple modules . * let @xmath7 be a connected gentle quiver with three vertices , with corresponding gentle algebra @xmath30 . by corollary [ orient - cycles ] we deduce that @xmath200 . algebras with different cartan determinant can not be derived equivalent , by theorem [ unimod ] . as an illustration , we shall give a complete derived equivalence classification of those algebras with cartan determinant 0 . by corollary [ det ] , a gentle algebra has cartan determinant 0 if and only if the quiver contains an even oriented cycle with full zero relations . there are 18 connected gentle quivers with three vertices having cartan determinant 0 , as listed in the following figure . 0.5 cm the main tool will be corollary [ ec - oc ] which states that the numbers @xmath33 and @xmath32 of minimal oriented cycles with full zero relations of even ( resp . odd ) length are invariants of the derived category . this will already settle large parts of the classification . in addition we will need to look at the centers and at the first hochschild cohomology group . the following table collects all the necessary invariants . again , in the cases where the quiver has loops , the dimension of @xmath193 depends on the characteristic being 2 or not , and in these cases the dimension in characteristic 2 is given in parantheses . @xmath201 for the derived equivalence classification , it only remains to consider those algebras having the same invariants . in the cases where the algebras are in fact derived equivalent , we leave out the details of the construction of a suitable tilting complex ; in the appendix a detailed example is provided which serves to indicate the strategy which also works in all other cases . the algebras @xmath202 and @xmath203 are derived equivalent . moreover , the algebras @xmath204 and @xmath205 are derived equivalent . note that @xmath202 and @xmath204 represent different derived equivalence classes since their first hochschild cohomology groups have different dimensions . the algebras @xmath206 and @xmath207 are derived equivalent . ( the details for this case are provided in the appendix . ) similarly , the algebras @xmath208 , @xmath209 , @xmath210 and @xmath211 are derived equivalent , the algebras @xmath212 and @xmath213 are derived equivalent and moreover , the algebras @xmath214 and @xmath215 are derived equivalent . the case of @xmath216 is more subtle . this algebra has exactly the same invariants as the algebras @xmath204 and @xmath205 . however , we claim that @xmath216 is not derived equivalent to @xmath205 . in fact , the lie algebra structures on @xmath193 are not isomorphic . note that with the gerstenhaber bracket , the first hochschild cohomology becomes a lie algebra . by a result of b. keller @xcite , this lie algebra structure on @xmath193 is invariant under derived equivalence . as mentioned before , by work of m. bardzell @xcite there is an explicit way of computing @xmath193 for a gentle algebra , and a nice combinatorial version due to c. strametz ( * ? ? ? * proposition 2.6 ) ( for the additive structure ) and ( * ? ? ? * theorem 2.7 ) ( for the lie algebra structure ) . with this method one can compute that the four - dimensional lie algebras on @xmath217 and on @xmath218 are not isomorphic . in fact , the lie algebra center of @xmath217 is two - dimensional , whereas the lie algebra center of @xmath218 has dimension 1 . this completes the derived equivalence classification of connected gentle algebras with three simple modules and cartan determinant 0 . the ten derived equivalence classes are indicated in the above table by the horizontal double lines . skewed - gentle algebras were introduced in @xcite ; for the notation and definition we follow here mostly @xcite , but we try to explain how the construction works rather than repeating the technical definition from @xcite . we start with a gentle pair @xmath7 . a set @xmath219 of vertices of the quiver @xmath11 is an admissible set of _ special _ vertices if the quiver with relations obtained from @xmath11 by adding loops with square zero at these vertices is again gentle ; we denote this gentle pair by @xmath220 . the triple @xmath221 is then called _ skewed - gentle_. we want to point out that the admissibility of the set @xmath219 of special vertices is both a local as well as a global condition . let @xmath54 be a vertex in the gentle quiver @xmath7 ; then we can only add a loop at @xmath54 if @xmath54 is of valency 1 or 0 or if it is of valency 2 with a zero relation , but not one coming from a loop . hence only vertices of this type are potential special vertices . but for the choice of an admissible set of special vertices we also have to take care of the global condition that after adding all loops , the pair @xmath220 still does not have paths of arbitrary lengths . given a skewed - gentle triple @xmath221 , we now construct a new quiver with relations @xmath222 by doubling the special vertices , introducing arrows to and from these vertices corresponding to the previous such arrows and replacing a previous zero relation at the vertices by a mesh relation . + more precisely , we proceed as follows . the non - special ( or : ordinary ) vertices in @xmath11 are also vertices in the new quiver ; any arrow between non - special vertices as well as corresponding relations are also kept . any special vertex @xmath223 is replaced by two vertices @xmath224 and @xmath225 in the new quiver . an arrow @xmath226 in @xmath11 from a non - special vertex @xmath160 to @xmath54 ( or from @xmath54 to @xmath160 ) will be doubled to arrows @xmath227 ( or @xmath228 , resp . ) in the new quiver ; an arrow between two special vertices @xmath229 will correspondingly give four arrows between the pairs @xmath230 and @xmath231 . we say that these new arrows lie over the arrow @xmath226 . any relation @xmath232 where @xmath233 is non - special gives a corresponding zero relation for paths of length 2 with the same start and end points lying over @xmath234 . if @xmath54 is a special vertex of valency 2 in @xmath11 , then the corresponding zero relation at @xmath54 , say @xmath232 with @xmath235 , is replaced by mesh commutation relations saying that any two paths of length 2 lying over @xmath234 , having the same start and end points but running over @xmath224 and @xmath225 , respectively , coincide in the factor algebra to the new quiver with relations @xmath222 . + we will speak of @xmath222 as a skewed - gentle quiver _ covering _ the gentle pair @xmath7 . a @xmath61-algebra is then called _ skewed - gentle _ if it is morita equivalent to a factor algebra @xmath236 , where @xmath237 comes from a skewed - gentle triple @xmath221 as above . * remark . * let @xmath30 be gentle . in a gentle quiver , there is at most one non - zero cyclic path starting and ending at a given vertex ; hence the diagonal entries in the @xmath0-cartan matrix @xmath24 are @xmath29 or of the form @xmath238 , for some @xmath239 . + if a vertex @xmath54 in @xmath11 can be chosen as a special vertex for a covering skewed - gentle quiver @xmath240 , then the corresponding diagonal entry in @xmath24 is @xmath29 , as otherwise we have paths of arbitrary lengths in @xmath241 ; hence in the corresponding @xmath0-cartan matrix for the skewed - gentle algebra @xmath242 we have @xmath243 on the diagonal for the two split vertices @xmath230 in @xmath240 . + [ thm : skewed - gentle ] let @xmath7 be a gentle quiver , and @xmath237 a covering skewed - gentle quiver . let @xmath41 be the corresponding skewed - gentle algebra . denote by @xmath23 the number of oriented @xmath12-cycles in @xmath7 with full zero relations . + then the @xmath0-cartan matrix @xmath42 is unimodularly equivalent ( over @xmath19 $ ] ) to a diagonal matrix with entries @xmath43 , with multiplicity @xmath27 , @xmath28 , and all further diagonal entries being @xmath29 . * again , we argue by induction on the number of vertices and arrows . we let @xmath30 be the gentle algebra and @xmath24 the @xmath0-cartan matrix as before . + if @xmath11 has no arrows , then @xmath240 is just obtained by doubling the special vertices , and this still has no arrows , so the result clearly holds . + if @xmath11 has an arrow @xmath123 as in lemma [ lem : arrow - removal ] , with a non - special @xmath244 in the first case , and a non - special @xmath245 in the second case , respectively , then we can argue as in the proof of lemma [ lem : arrow - removal ] to remove @xmath123 . let us consider again the situation of the first case , so here @xmath246 is non - special . note that a maximal non - zero path @xmath13 starting from @xmath135 with @xmath123 or @xmath247 ( if @xmath248 is special ) will end on a non - special vertex ( and hence this maximal path is unique in @xmath242 ) ; in general , this path will be longer than the one taken in @xmath4 . in the column transformations , we only have to be careful at doubled vertices on the path @xmath13 ; here we replace both corresponding columns @xmath249 of @xmath42 by @xmath250 . this leads to the cartan matrix for the skewed - gentle algebra where @xmath123 or @xmath247 , respectively , has been removed from @xmath240 , which is a skewed - gentle cover for the quiver obtained from @xmath11 by deleting @xmath123 ; then the claim follows by induction . now assume @xmath11 has a source @xmath54 which is special ( w.l.o.g . the first vertex ) ; the case of a sink is dual . then the @xmath0-cartan matrix for @xmath242 has the form @xmath251 where @xmath252 is the @xmath0-cartan matrix of the skewed - gentle algebra @xmath253 for the quiver @xmath254 obtained from @xmath240 by removing @xmath255 and the arrows incident with @xmath230 . note that @xmath254 is the skewed - gentle cover for the quiver @xmath120 which is obtained from @xmath11 by removing @xmath54 and the arrow incident with @xmath54 , and the choice @xmath256 as the set of special vertices ; in short , we write this as @xmath257 . again , using induction the claim follows immediately . thus again , we may now assume that @xmath11 has only vertices of valency 2 with a ( non - loop ) zero relation or vertices of valency 4 ; note that any special vertex in @xmath11 has to be of valency 2 . as before , we may also assume that @xmath11 ( and hence also @xmath240 ) are connected . if there are no non - special vertices , or if all non - special vertices are of valency 4 , then @xmath241 is not gentle . hence @xmath11 has a non - special vertex @xmath54 of valency 2 with a zero relation at @xmath54 . let @xmath258 be the ( unique ) incoming arrow . + again we consider the unique maximal path @xmath13 in @xmath11 with non - repeating arrows starting in @xmath135 with @xmath126 , such that the product of any two consecutive arrows is zero in @xmath4 ; as before , we note that in our current situation @xmath13 has to be a cycle @xmath259 , where @xmath128 , @xmath155 , and @xmath260 . as in the previous situation , we note that the arrows are distinct , but vertices @xmath131 may be repeated . + for a vertex @xmath160 in @xmath11 we denote by @xmath159 the row of the @xmath0-cartan matrix @xmath24 corresponding to @xmath160 . if @xmath160 is non - special , we denote by @xmath261 the corresponding row in the @xmath0-cartan matrix @xmath262 . if @xmath160 is special , then for the two vertices @xmath231 we have two corresponding rows @xmath263 in the cartan matrix @xmath264 , and we then set @xmath265 . + before , we have transformed @xmath111 by replacing @xmath161 by @xmath266 and obtained a matrix @xmath146 . we now do a parallel transformation on @xmath264 , that is , we replace @xmath267 by @xmath268 and we obtain a matrix @xmath269 . we have to compare the differences and check that everything stays under control for the induction argument . + if a vertex @xmath137 , @xmath270 , is special , note that the doubled contribution in @xmath271 is needed on the one hand for the cancellation with the previous row , and on the other hand to continue around the cycle @xmath153 . as @xmath54 is non - special and of valency 2 with a zero - relation , we note that as before , in @xmath272 we only have the contribution @xmath175 at @xmath54 . following this by the parallel operation to the previous column operation we then replace the column @xmath273 by the linear combination @xmath274 where we use analogous conventions as before . with @xmath54 corresponding to the first row and column of the cartan matrix , we have thus unimodularly transformed @xmath264 to a matrix of the form @xmath275 where @xmath252 is the cartan matrix of the skewed - gentle algebra @xmath253 for the quiver @xmath254 obtained from @xmath240 by removing @xmath54 and the arrows incident with @xmath54 . note that in fact , @xmath276 in the notation of our previous proof , i.e. , as explained earlier , @xmath254 is the skewed - gentle cover for the quiver @xmath120 and the choice @xmath256 as the set of special vertices . thus the result follows by induction . @xmath147 + by comparing theorem [ thm : gentle ] and theorem [ thm : skewed - gentle ] we observe that the @xmath0-cartan matrix @xmath24 for the gentle algebra @xmath4 to @xmath7 , and the @xmath0-cartan matrix @xmath42 for a skewed - gentle cover @xmath242 are unimodularly equivalent to diagonal matrices which only differ by adding as many further @xmath29 s on the diagonal as there are special vertices chosen in @xmath11 . in particular , with notation as above , @xmath277 this observation has the following immediate consequence when specializing to @xmath1 . let @xmath7 be a gentle quiver , and @xmath237 a covering skewed - gentle quiver . then the determinant of the ordinary cartan matrix of the skewed - gentle algebra @xmath41 is the same as the one for the gentle algebra @xmath30 , i.e. , @xmath45 . a gentle algebra and a ( proper ) skewed - gentle algebra may have the same @xmath0-invariants but they can not be derived equivalent by @xcite , corollary 1.2 . this appendix is aimed at providing enough background on tilting complexes and explicit computations of their endomorphism rings so that the interested reader can fill in the details in the derived equivalence classifications of examples [ derived - two ] and [ derived - three ] . we explained there in detail how to distinguish derived equivalence classes ( since this is the main topic of this paper ) , but have been fairly short on indicating why certain algebras in the lists are actually derived equivalent . in this section we will go through one example in detail ; this will indicate the main strategy which also works in all other cases . two algebras @xmath4 and @xmath94 are called derived equivalent if @xmath278 and @xmath280 are equivalent as triangulated categories . by j. rickard s theorem @xcite , this happens if and only there exists a tilting complex @xmath281 for @xmath4 such that the endomorphism ring @xmath282 in the homotopy category is isomorphic to @xmath94 . a bounded complex @xmath281 of projective @xmath4-modules is called a tilting complex if the following conditions are satisfied . recall from section [ sec - def ] our conventions to deal with right modules and to read paths from left to right . in particular , left multiplication by a nonzero path from vertex @xmath79 to vertex @xmath78 gives a homomorphism @xmath287 . we define the following bounded complex @xmath288 of projective @xmath206-modules . let @xmath289 and @xmath290 be stalk complexes concentrated in degree 0 . moreover , let @xmath291 ( in degrees 0 and @xmath292 ) . we claim that @xmath281 is a tilting complex . property ( i ) above is obvious for all @xmath293 since we are dealing with two - term complexes . let @xmath294 , and consider possible maps @xmath295 $ ] where @xmath296 . this is given by a map of complexes as follows @xmath297 where @xmath11 could be either of @xmath298 , @xmath299 , or @xmath300 . but since we are dealing with gentle algebras , no nonzero map can be zero when composed with both @xmath100 and @xmath99 . so the only homomorphism of complexes @xmath295 $ ] is the zero map , as desired . directly from the definition we see that @xmath301)=0 $ ] and @xmath302)=0 $ ] ( since they are stalk complexes ) . now let @xmath304 . we have to consider maps @xmath305 $ ] ; these are given as follows @xmath306 where @xmath11 again can be either of @xmath298 , @xmath299 , or @xmath300 . now there certainly exist nonzero homomorphisms of complexes . but they are all homotopic to zero . in fact , every path in the quiver of @xmath206 from vertex 2 to vertex 1 or 3 either starts with @xmath100 or with @xmath99 . accordingly , every homomorphism @xmath307 can be factored through the map @xmath308 . it remains to show that the complex @xmath281 also satisfies property ( ii ) of the definition of a tilting complex . it suffices to show that the projective indecomposable modules @xmath298 , @xmath310 and @xmath299 , viewed as stalk complexes , can be generated by @xmath286 . this is clear for @xmath298 and @xmath299 since they occur as summands of @xmath281 . for @xmath310 , consider the map of complexes @xmath311 given by the identity map on @xmath300 in degree 0 . then the stalk complex @xmath312 $ ] with @xmath310 in degree 0 can be shown to be homotopy equivalent ( i.e. isomorphic in @xmath279 ) to the mapping cone of @xmath313 . thus we have a distinguished triangle @xmath314 \to \underbrace{t_2[1]}_{\in{\operatorname{add}}(t)}.\ ] ] by definition , @xmath286 is triangulated , so it follows that also the stalk complex @xmath312\in{\operatorname{add}}(t)$ ] , which proves ( ii ) . by rickard s theorem , the endomorphism ring of @xmath281 in the homotopy category is derived equivalent to @xmath206 . we need to show that @xmath315 is isomorphic to @xmath207 . note that the vertices of the quiver of @xmath316 correspond to the summands of @xmath281 . _ alternating sum formula . _ for a finite - dimensional algebra @xmath4 , let @xmath317 and @xmath318 be bounded complexes of projective @xmath4-modules . then @xmath319)= \sum_{r , s } ( -1)^{r - s } \dim{\operatorname{hom}}_a(q^r , r^s).\ ] ] in particular , if @xmath11 and @xmath320 are direct summands of a tilting complex then @xmath321 note that the cartan matrix of @xmath206 has the form . from that , using the alternating sum formula , we can compute the cartan matrix of @xmath316 to be . note that this is actually the cartan matrix of @xmath207 . now we have to define maps of complexes between the summands of @xmath281 , corresponding to the arrows of the quiver of @xmath207 . the final step then is to show that these maps satisfy the defining relations of @xmath207 , up to homotopy . we define @xmath322 by the map @xmath323 in degree 0 . note that this is indeed a homomorphism of complexes since @xmath324 in @xmath206 . moreover , we define @xmath325 and @xmath326 by the projection onto the first and second summand in degree 0 , respectively . finally , we define @xmath327 by @xmath328 . we now have to check the relations , up to homotopy . we write compositions from left to right ( as in the relations of the quiver of @xmath316 ) . clearly , @xmath329 . the composition @xmath330 is given in degree 0 by @xmath331 . so it is not the zero map , but is homotopic to zero via the homotopy map @xmath332 ( use that @xmath333 in @xmath206 ) . finally , consider @xmath334 on @xmath335 . it is given by @xmath336 in degree 0 and the zero map in degree @xmath292 . it is indeed homotopic to zero via the homotopy map @xmath337 . ( note that here we use that @xmath338 and @xmath324 in @xmath206 . ) thus , we have defined maps between the summands of @xmath281 , corresponding to the arrows of the quiver of @xmath207 . we have shown that they satisfy the defining relations of @xmath207 , and that the cartan matrices of @xmath316 and @xmath207 coincide . from this we can conclude that @xmath339 . hence , @xmath206 and @xmath207 are derived equivalent , as desired . all the other derived equivalences stated in examples [ derived - two ] and [ derived - three ] can be verified exactly along these lines . in particular , they can also be realized by tilting complexes with non - zero entries in only two degrees .	cartan matrices are of fundamental importance in representation theory . for algebras defined by quivers with monomial relations the computation of the entries of the cartan matrix amounts to counting nonzero paths in the quivers , leading naturally to a combinatorial setting . our main motivation are derived module categories and their invariants : the invariant factors , and hence the determinant , of the cartan matrix are preserved by derived equivalences . the paper deals with the class of ( skewed- ) gentle algebras which occur naturally in representation theory , especially in the context of derived categories . we study @xmath0-cartan matrices , where each nonzero path is weighted by a power of an indeterminate @xmath0 according to its length . specializing @xmath1 gives the classical cartan matrix . we determine normal forms for the @xmath0-cartan matrices of skewed - gentle algebras . in particular , we give explicit combinatorial formulae for the invariant factors and thus also for the determinant . as an application of our main results we show how to use our formulae for the difficult problem of distinguishing derived equivalence classes . msc - classification : 16g10 , 18e30 , 05e99 , 05c38 , 05c50 * @xmath0-cartan matrices and combinatorial invariants of derived categories for skewed - gentle algebras * * christine bessenrodt * fakultt fr mathematik und physik , universitt hannover , + welfengarten 1 , d-30167 hannover , germany * thorsten holm * department of pure mathematics , university of leeds , + leeds ls2 9jt , england
this work is financially supported by a grant - in - aid for specially promoted research ( no . 20001006 ) from the ministry of education , culture , sports , science and technology .	standing waves near the zigzag and armchair edges , and their berry s phases are investigated . it is suggested that the berry s phase for the standing wave near the zigzag edge is trivial , while that near the armchair edge is non - trivial . a non - trivial berry s phase implies the presence of a singularity in parameter space . we have confirmed that the dirac singularity is absent ( present ) in the parameter space for the standing wave near the zigzag ( armchair ) edge . the absence of the dirac singularity has a direct consequence in the local density of states near the zigzag edge . the transport properties of graphene nanoribbons observed by recent numerical simulations and experiments are discussed from the point of view of the berry s phases for the standing waves . the electronic property of a graphene nanoribbon differs greatly from that of a carbon nanotube . a metallic carbon nanotube exhibits a high mobility , while a graphene nanoribbon shows a transport gap . @xcite the high mobility observed in metallic carbon nanotubes indicates that the scatterers are not effective in producing backward scattering . there should be a mechanism which suppresses the backward scattering in metallic carbon nanotubes . suppose that an electron with momentum @xmath0 is coming into the impurities which are represented by the circles in fig . [ fig : abs](a ) . the electron is scattered by the impurities and changes its momentum direction . let us consider the probability amplitude that the electron is scattered in the backward direction , as shown in fig . [ fig : abs](a ) . in this case , the final wave function , @xmath1 , is given by rotating the wave vector of the initial state , @xmath2 , by @xmath3 , so that we have the relationship between the initial state and final state as @xmath4 where @xmath5 is a rotational operator with angle @xmath6 . note that this particular path shown in fig . [ fig : abs](a ) is not the unique path that an electron can follow . there is an another path that an electron can follow , which we denote it by the lines in fig . [ fig : abs](b ) . the new path relates to the original path through the `` time reversal '' . we denote the final state in this `` time reversal '' path by @xmath7 . because the final state is given by rotating the wave vector by @xmath8 , we have the relationship , @xmath9 between the initial state and the final state . by eliminating the wave function of the initial state from eqs . ( [ eq : path1 ] ) and ( [ eq : path2 ] ) , we get @xmath10 now , the total backward scattering amplitude is given by the sum of @xmath1 and @xmath11 as @xmath12 \phi_{\bf -k}$ ] . because the wave function gets an extra phase shift of @xmath3 ( called the berry s phase @xcite ) through a rotation of the wave vector around the dirac point , that is , @xmath13 is equivalent to @xmath14 ( @xmath15 ) , a `` time - reversal '' pair of backward scattered waves cancels with each other , i.e. , @xmath16 @xcite . , combines with the final state of the scattering path shown in ( b ) , @xmath11 , to suppress the total backscattering amplitude , @xmath17 , due to the berry s phase . the two scattering paths , ( a ) @xmath18 and ( b ) @xmath19 , are related with each other by `` time - reversal '' . ] this , absence of backward scattering mechanism provides us with a simple solution explaining the high mobility , and the existence of a singularity at the dirac point is essential to the nontrivial phase shift of @xmath20 . however , it is not obvious whether the wave functions realizing in graphene nanoribbons can acquire a nontrivial berry s phase or not . the eigen state near the edge is the standing wave resulting from the interference between an incident wave and the edge reflected wave . if the berry s phase for the standing wave is trivial , then the scatters , such as a potential disorder created by charge impurities , may give rise to backward scattering . as a result , the mobility of a graphene nanoribbon decreases considerably than that of a carbon nanotube . @xcite in the present paper , we study the berry s phases of the standing waves near the zigzag and armchair edges , and their effects on the local density of states and the transport properties of graphene nanoribbons . in fig . [ fig : unit ] , we consider the zigzag edge parallel to the @xmath21-axis , by which translational symmetry along the @xmath22-axis is broken . thus , the incident state with wave vector @xmath23 is elastically scattered by the zigzag edge , and the wave vector of the reflected state becomes @xmath24 . by contrast , the armchair edge parallel to the @xmath22-axis breaks translational symmetry along the @xmath21-axis , so that the wave vector of the reflected state is @xmath25 . since the brillouin zone ( bz ) is given by rotating the hexagonal lattice by 90@xmath26 , one can see in fig . [ fig : unit ] that for the incident state near the k point , the reflected state by the zigzag edge is also near the k point , while the reflected state by the armchair edge is near the k@xmath27 point . hence , the scattering by the zigzag edge is intravalley scattering , while that by the armchair edge is intervalley scattering . coordinate system is fixed as shown . ( right ) the hexagon represents the first bz of graphene , and the corners of the hexagon are k and k@xmath27 points . ] to begin with , let us consider the scattering problem for the zigzag edge . since the zigzag edge is not the source of intervalley scattering , we focus on only the electrons near the k point . the incident and reflected waves are represented by the bloch states . the bloch state in the conduction energy band is written as @xmath28 where @xmath29 is the area of sample , @xmath0 is the momentum measured from the k point , and @xmath30 is the polar angle between the vector @xmath0 and the @xmath31-axis . note that the bloch function acquires a nontrivial berry s phase of @xmath3 by a rotation of the wave vector around the k point as @xmath32 the standing wave modes arise from the interference between the incident waves and the reflected waves . let the wave vector of the incident wave be @xmath33 . then the vector of the elastically reflected wave is given by @xmath34 , and the standing wave near the zigzag edge is written as @xmath35 where the phase @xmath36 should be determined in such a manner that @xmath37 [ @xmath38 satisfies the boundary condition @xcite . we take the following boundary condition for the zigzag edge @xcite , @xmath39 which describes a situation in which the b - atoms located slightly above the horizontal dashed line at @xmath40 in fig . [ fig : unit ] are separated from the lower semi - infinite graphene for @xmath41 . the boundary condition of eq . ( [ eq : phib0 ] ) is satisfied when @xmath42 , and the standing wave is written as @xmath43 where @xmath44 is the length of the zigzag edge , and @xmath45 for @xmath46 and @xmath47 otherwise . the value of @xmath48 is determined by the normalization condition . we note that the standing wave of eq . ( [ eq : zigwfsum ] ) reproduces the result of tight - binding lattice model ( see appendix b of ref . ) . first , we examine the berry s phase of the standing wave . for simplicity , let us consider the case @xmath49 for eq . ( [ eq : s - sol ] ) . this corresponds to a normal incident process to the zigzag edge [ @xmath50 and @xmath51 . the amplitude for this process does not vanish as @xmath52 this standing wave includes a backward scattering amplitude , and the existence of the backward scattering amplitude indicates that the berry phase of the standing wave vanishes . indeed , since @xmath53 , the berry s phase of the reflected state is given by @xmath8 as @xmath54 while that of the incident state is @xmath3 [ see eq . ( [ eq : berry- ] ) ] . the berry s phase of @xmath3 for the incident state is canceled by the berry s phase of @xmath20 for the reflected state , and hence the berry s phase for the standing wave vanishes in total . the trivial berry s phase can be understood from eq . ( [ eq : s - sol ] ) , in which the bloch function of the standing wave is real @xcite . next , we study the standing wave near the armchair edge . the k and k@xmath27 points need to be considered simultaneously in the case of armchair edge since the armchair edge is the source of intervalley scattering . suppose that the wave vector of the incident wave measured from the k point is @xmath33 . thus , the wave vector of the elastically reflected wave measured from the k@xmath27 point is given by @xmath55 , and the standing wave is written as @xmath56 it can be shown that the bloch state near the k@xmath27 point is given by @xmath57 @xcite . thus , we may write @xmath58 where @xmath59 for @xmath60 and @xmath61 otherwise . the relative phase @xmath62 can be determined by the boundary condition for armchair edge . the value of @xmath62 must be @xmath63 , which will be derived elsewhere . @xcite note that the bloch functions for the k and k@xmath27 points are the same . thus , the berry s phase of the standing wave near the armchair edge is given by @xmath3 . as a result , we can expect that the absence of backward scattering mechanism is valid near the armchair edge . note , however , that the absence of backward scattering mechanism for the armchair edge is not identical to that discussed so far for carbon nanotubes . this is because the notion of a `` time - reversal pair '' of scattered waves in the original argument @xcite should be replaced to a true time - reversal pair of scattered waves near the armchair edge . the time - reversal state of eq . ( [ eq : armwf ] ) is given by @xcite @xmath64 a magnetic field breaks time - reversal symmetry , so that it invalidates the new absence of backward scattering mechanism and leads to weak anti - localization . it is also interesting to point out that the berry s phase for the standing wave near the armchair edge is robust against an ordinary mass term , while it is not robust against a topological mass term @xcite . the appearance of a nontrivial berry s phase is related to a singularity in parameter space @xcite . here , we show the relationship between the singularity at the dirac point ( @xmath65 ) in the momentum space and the standing wave . for the zigzag edge , @xmath31 is a continuous variable , and therefore the spectrum with @xmath66 may cross the dirac singularity . however , eq . ( [ eq : s - sol ] ) shows that the standing wave with @xmath67 does not exist , that is @xmath68 . thus , the dirac singularity seems to be separated from the spectrum of the standing wave . to explore this point further , we consider a zigzag nanoribbon by introducing another zigzag edge at @xmath69 , in addition to the zigzag edge at @xmath40 . suppose that the edge atoms at @xmath69 are b - atoms , which imposes the boundary condition on the wave function at @xmath70 as @xmath71 . this leads to the constraint equation for @xmath23 , @xmath72 where @xmath73 is an integer . we note that this equation provides the solutions of @xmath74 which was obtained by brey and fertig in ref . [ the minus sign in front of @xmath31 is a matter of notation ] . note that @xmath73 must be a nonzero integer because the equation does not possess a solution when @xmath75 . in fig . [ fig : sflow ] , we plot the solutions of eq . ( [ eq : const ] ) for the cases @xmath76 and 2 . the solution for @xmath77 is given by shifting the curve for @xmath78 by @xmath79 , and only the solution with @xmath80 shows an anomalous feature . the curve with @xmath80 approaches a point on the @xmath31-axis . this behavior can be checked by setting @xmath81 in eq . ( [ eq : const ] ) with @xmath80 . it is appropriate to refer to this state with @xmath82 and @xmath66 as the critical state @xcite because this state is neither the standing wave nor the edge state @xcite . the fact that no curve crosses the dirac singularity is consistent to the absence of a nontrivial berry s phase for the standing wave . in other words , an electron can not approach the dirac singularity due to the presence of the zigzag edge . as a result , an energy gap appears in the spectrum of the standing wave . the minimum energy gap is determined by the critical state as @xmath83 where @xmath84 is the fermi velocity . in contrast to the boundary condition for the zigzag edge , the boundary condition for the armchair edge does not forbid an electronic state cross the dirac singularity point , and therefore the electron can pick up a nontrivial berry s phase . and @xmath85 which are allowed by the boundary condition for a zigzag nanoribbon . the dirac singularity is not located on the curves , which shows that there is no singularity at the parameter space of the standing waves . the inset shows the corresponding spectral flow in the energy dispersion relation . ] here , we consider pseudospin of the standing wave in order to examine the local density of states ( ldos ) near the zigzag edge . the pseudospin for an eigenstate @xmath86 is defined by the expectation value of pauli matrices as @xmath87 ( @xmath88 ) where @xmath89 is a pseudospin density defined by @xmath90 . the boundary condition of eq . ( [ eq : phib0 ] ) means that the pseudospin density is polarized into the positive @xmath91-axis locally near the zigzag edge , that is , @xmath92 and @xmath93 . actually , by putting @xmath94 into eq . ( [ eq : s - sol ] ) , we see that the standing wave near the zigzag edge has amplitudes only on a - atoms . this polarization of the pseudospin causes an anomalous behavior to appear in the ldos near the zigzag edge . to show this , let us first review the ldos for a graphene without an edge . assuming that electrons are non - interacting , the bulk ldos is given by @xmath95 where @xmath96 is proportional to @xmath97 , which results from the dirac cone spectrum . the ldos near the zigzag edge can be calculated by using the standing wave given in eq . ( [ eq : s - sol ] ) . the ldos has the form , @xmath98 where @xmath99 is defined as @xmath100 by performing the integral with respect to the angle @xmath6 in eq . ( [ eq : rfac ] ) , we obtain the analytical result for @xmath99 as @xmath101 where @xmath102 is a bessel function of order @xmath103 . the ldos is symmetric with respect to @xmath104 . the bulk ldos in eq . ( [ eq : ldos_bulk ] ) can be reproduced by setting @xmath105 in eq . ( [ eq : ldos ] ) since @xmath106 . in fig . [ fig : pspin ] , we plot the ldos at @xmath107 , 1 , 2 , 3 , and 6[nm ] . the slope of the ldos depends on the distance from the zigzag edge . at the edge , that is , at @xmath107 , since @xmath108 and @xmath109 , we see that @xmath110 , and the slope of @xmath111 is half of that of @xmath112 [ @xmath113 . this results from the fact that the amplitudes at one of two sublattice disappears near the zigzag edge ( due to the pseudospin polarization ) , and only the half of the amplitude in the unit cell can contribute to the ldos . is due to the edge states . ] in fig . [ fig : pspin ] , in addition to the ldos due to the standing wave , we plot the ldos ( at @xmath114[nm ] ) due to the edge states @xcite . the edge states produce a peak at @xmath104 in the ldos . the ldos of the edge states is calculated as follows . the wave function of the edge state is given by @xcite @xmath115 note that the edge states appear for @xmath116 and @xmath41 ( see fig . [ fig : sflow ] ) . since the energy eigenvalue of the edge state vanishes , the ldos for the edge states is written as @xmath117 where @xmath118 is a phenomenological parameter representing energy uncertainty of the edge states , for which we assume @xmath119 mev . substituting @xmath120 with @xmath121 in eq . ( [ eq : rhoes ] ) , and performing the integral for @xmath31 , we obtain @xmath122 this result has been used in fig . [ fig : pspin ] to plot the ldos of the edge states . note that @xmath123 decreases as @xmath124 , which is a slowly decreasing function compared with the exponential decaying wave function of the edge state . it is interesting to note that @xmath125 can be derived from the fact that @xmath126 does not depend on @xmath22 for a large value of @xmath127 . this is a sum rule for the total ldos consisting of the edge states and the standing waves , in which the density above ( below ) the fermi energy at @xmath104 must be position independent . this argument also shows that the singular behavior of the ldos for the edge states at @xmath40 is a consequence of the wave function of the standing wave . by examining the standing wave near the graphene edge , we have seen that the armchair edge preserves a non - trivial berry s phase as in the bulk of graphene , on the other hand , such a non - trivial phase is absent for the standing wave near the zigzag edge . here , we consider the indication of this result with respect to the transport in graphene nanoribbons . in the argument for the absence of the backward scattering given in the introduction ( see fig . [ fig : abs ] ) , we find , by replacing @xmath2 with the standing wave @xmath128 , that the backscattering amplitude , @xmath129\psi_{\bf -k}$ ] , is enhanced for zigzag nanoribbons ( due to @xmath130 ) , while it still vanishes for armchair nanoribbons ( due to @xmath131 ) . thus , we can expect that the conductance of a zigzag nanoribbon is smaller than that of an armchair nanoribbon , if we assume only scatters with long - range potential . the simulations performed by areshkin _ et al . _ @xcite and yamamoto _ et al . _ @xcite show indeed that long - range potentials are ineffective in causing backscattering in a perfect armchair nanoribbon . for narrow metallic armchair nanoribbons , the linear energy dispersion originating from the dirac point provides a nearly perfect conduction , and this nearly perfect conduction might relate to the nontrivial berry s phase since the linear energy dispersion picks up the dirac singularity . note , however , that the standing wave for the armchair edge is a short - length intervalley mode , that is , the electronic wave length of the standing wave is order of the lattice constant . as a result , the wave function can be very sensitive to a short - range scattering potential , such as irregular edges @xcite and lattice vacancies @xcite . an armchair nanoribbon has both advantages ( non - trivial berry s phase ) and disadvantages ( intervalley mode ) for the transport as well as that a zigzag nanoribbon has both advantages ( intravalley mode ) and disadvantages ( trivial berry s phase ) . therefore , in a realistic situation , it is reasonable to consider that a nanoribbon does not exhibit an electronic conduction comparable to a metallic carbon nanotube . it seems that recent tight - binding numerical studies show that the conductance of zigzag nanoribbons is more robust than that of armchair nanoribbons against edge disorders . @xcite in the numerical studies , the transport given by the electrons which are located very close to the fermi level is concerned . for zigzag nanoribbons , the propagating modes in each valley contain a single one - way excess channel ( the states on the curve of @xmath80 which are close to the critical state shown in fig . [ fig : sflow ] ) . this feature causes a perfectly conducting channel to appear in the disordered system for the case that impurities do not give rise to intervalley scattering @xcite . this feature also seems to cause a robust conducting channel to appear in the edge disordered system . @xcite for our discussion of the backscattering in zigzag nanoribbons , we assume that the initial state with momentum @xmath31 accompanies the corresponding final state with @xmath132 . this condition is not satisfied for the single one - way excess channel , for which our discussion for the berry s phase is useless . the single one - way excess channel plays a dominant role in zigzag nanoribbons of the widths up to several nanometers for a limited energy window . @xcite our result for the berry s phase may be useful in discussing transport of zigzag nanoribbons of the widths more than 10 nm , for which we have several number of channels having the final state with @xmath132 . our result may have a relationship to the formation of a transport gap observed for graphene nanoribbons . @xcite note that the standing wave near a realistic rough edge can be described as the sum of the standing waves for zigzag and armchair edges . @xcite since the standing wave for zigzag edge is a long - length intravalley mode and that for armchair edge is a short - length intervalley one , their characters do not interfere with each other and can coexist in the standing wave near a rough edge . this is also indicative of that a small portion of zigzag edge in a rough edge eventually governs the long - length transport behavior of a nanoribbon , as is numerically simulated by several authors . @xcite indeed , gallagher _ et al . _ @xcite and han _ et al . _ @xcite observed that the source - drain gap has a strong dependence on ribbon s length . it is also important to recognize that there are two distinct localized states ; the edge states and localized states which consist of the standing waves . since the localization length of the former is on the order of lattice constant , it is difficult to consider that the transport behavior of nanoribbons of the widths larger than @xmath133 nm is dominated by the edge states . @xcite in contrast , the latter is caused by the scatters , such as a potential disorder created by charge impurities , giving rise to backward scattering . since the latter consists of the standing waves , there is a possibility that the localization length is the order of the width of a ribbon . han _ et al . _ @xcite measured electron transport in lithographically fabricated nanoribbons of widths @xmath134 nm with rough edge on the order of nanometers , and confirmed that the size of a transport gap is inversely proportional to the ribbon width . in conclusion , we have shown that the berry s phase for the standing wave near the zigzag edge is trivial . the momentum parameter space for the standing wave near the zigzag edge does not include the dirac singularity , which is essential to the absence of a non - trivial berry s phase . as a result , the absence of backward scattering mechanism that works well in a metallic carbon nanotube can not be used for a zigzag nanoribbon . the absence of the dirac singularity in the momentum space of the standing wave results in the peak in the ldos due to the edge states . an observation of the ldos peak near the zigzag edge is a direct evidence of the absence of the dirac singularity in the parameter space of the standing wave . in contrast , a non - trivial berry s phase survives the reflection by the armchair edge . for the standing wave near the armchair edge , we obtain the new absence of backward scattering mechanism in which a real time - reversal pair of backward scattered waves cancels with each other . this absence of the backward scattering mechanism for the armchair edge is not robust against a short - range scattering potential since the standing wave is a short - length intervalley mode . an armchair nanoribbon has both advantage and disadvantage for the transport as well as a zigzag nanoribbon .
[ intro ] experimental searches for the lepton - number violating weak process neutrinoless double beta decay ( @xmath0 decay ) are the most promising approach to determine some of the fundamental properties of neutrinos . the detection of @xmath0 decay would establish the majorana character of neutrinos and provide information about their absolute mass and hierarchy @xcite . ongoing experiments exo @xcite , kamland - zen @xcite and gerda @xcite have recently set impressive lower - limits , well over @xmath4 years , on the half - lives of @xmath5xe and @xmath6ge , and the dozen most favorable nuclei are being explored worldwide @xcite . however , an eventual @xmath0 decay detection does not guarantee the precise determination of absolute neutrino masses , because the half - life depends on the transition nuclear matrix elements ( nme ) , @xmath7 @xcite : @xmath8^{-1}=g_{0\nu}\|m^{0\nu}\|^{2}\left(\frac{\langle m_{\beta\beta}\rangle}{m_{e}}\right)^{2 } , \label{half_life}\ ] ] with @xmath9 the electron mass and @xmath10 a well - known phase - space factor @xcite . the combination of neutrino masses that appears in @xmath0 decay is @xmath11 , with @xmath12 the neutrino mixing matrix . nmes have been predicted by different theoretical nuclear structure approaches . these comprise large - scale shell model ( sm ) @xcite , energy density functional methods ( edf ) @xcite , the quasiparticle random phase approximation ( qrpa ) @xcite , and the interacting boson model ( ibm ) @xcite . however , state - of - the - art nme predictions by these approaches differ up to factor two @xcite , strongly limiting the precision to which information on neutrino masses would be known in case of a @xmath0 decay measurement . in addition , the half - lives of experimentally relevant nuclei could be significantly under / overestimated for a given @xmath13 . a better understanding of the nmes is therefore crucial . recently , experimental observables relevant for @xmath0 decay have been measured @xcite , allowing checks for the nuclear structure methods @xcite , that in some cases resulted in closer nmes between different theoretical approaches @xcite . in this article we follow a complementary approach , studying non - physical @xmath0 decays along isotopic chains in the @xmath1-shell , with sm and edf methods . although within this region only @xmath2ca is an actual @xmath0 decay candidate , a comparison between methods is better established with systematic calculations . for instance , systematic edf calculations for the cadmium isotopes provided a better understanding of the role of deformation , pairing and shell effects in @xmath0 decay @xcite . systematic studies allow to analyze not only numerical values , but also trends , which are useful to identify similarities and differences between sm and edf . understanding these is essential to reduce the theoretical uncertainty in the @xmath0 decay nmes . [ theo_frame ] here we briefly describe the edf and sm calculations performed in this work , as well as the @xmath0 decay transition operator employed . a more extensive description of edf calculations can be found in ref . @xcite and references therein . sm details can be found in refs @xcite . the transition operator is discussed in depth in refs . @xcite . the initial and final edf states are found using the gogny d1s functional @xcite . beyond mean field effects such as particle number and angular momentum restoration are included in addition to axial quadrupole configuration mixing within the generator coordinate method ( gcm ) @xcite . this method has been extensively used to study several nuclear structure properties throughout the whole nuclear chart ( see for instance ref . @xcite for its global performance with an approximate gcm method ) . one of the advantages of this approach is the explicit calculation of the nmes as a function of the quadrupole deformation of initial and final nuclei . hence , full shape - mixing and spherical nmes can be compared . additional degrees of freedom such as fluctuations in the pairing field ( already applied to nmes in ref . @xcite ) , triaxiality or octupolarity are neglected due to prohibitive computational times . nevertheless , we do not expect from these improvements any qualitative difference with respect to the analysis presented in this work . sm calculations are performed in the valence space comprised by the 0@xmath14 , 1@xmath15 , 1@xmath16 and 0@xmath17 orbitals ( @xmath1 shell ) , on top of a @xmath18ca core . we employ the well - known kb3 g @xcite and gxpf1a @xcite effective interactions , which have been shown to describe well the nuclear structure of @xmath1 shell nuclei @xcite . while the configuration space is significantly smaller in sm than in edf calculations , the main advantage of the sm approach is that all possible correlations within this space are included . the effect of the reduced sm valence space has been recently studied in the framework of many - body perturbation theory @xcite , with moderate increases in the nmes . here we neglect these corrections . truncated calculations can be performed limiting the number of neutrons and protons not coupled in @xmath19 pairs ( seniority truncations ) @xcite , enabling the study of the correlations associated to high - seniority components . in addition , sm states obtained in the full @xmath1 shell have good isospin quantum number , and projection to good isospin can be performed for truncated calculations . the sm code nathan @xcite has been used throughout this work . with the initial and final states obtained with the edf and sm methods , we calculate the @xmath0 decay nmes as described in refs . these can be decomposed according to spin structure into three different terms @xmath20 + where @xmath21 and @xmath22 are the vector and axial coupling constants , respectively . the gamow - teller ( gt ) part , @xmath23 , is dominant , and the fermi ( f ) component , @xmath24 , accounts to @xmath25 of @xmath23 @xcite . the tensor contribution , @xmath26 , gives a very small correction @xcite . here we focus on the main nme component @xmath23 , and discuss @xmath24 in the context of isospin conservation . we assume the closure approximation , which has been shown to be a good approximation ( up to 10% ) @xcite , sufficient for the purposes of this work . in this scheme the gt and f parts of the nmes follow the transition operator evaluated between the initial and final states : @xmath27 + with @xmath28 + where @xmath29 is the isospin - lowering operator that transforms neutrons into protons , and @xmath30 are the pauli spin matrices . the neutrino potentials @xmath31 depend on the relative distance between the two decaying nucleons , @xmath32 , the mass number @xmath33 , and the closure energy @xmath34 . here we use @xmath35 mev for all decays , the standard value used for @xmath2ca @xcite . a detailed form of the neutrino potentials can be found in refs . @xcite . in addition , short - range correlations are included within the ucom framework @xcite . other prescriptions have been recently proposed @xcite , but the differences are small , and not relevant for the purpose of this work . here we neglect two - body currents , related to the effective quenching of the @xmath36 operator @xcite in weak decays , and restrict to a purely two - body operator derived from one - body currents only . while two - body current contributions may be important for the absolute value of the nmes @xcite , including them would not alter the main conclusions of this study . [ results ] with the transition operator described in sec . [ theo_frame ] , identical for sm and edf calculations , we can make a direct comparison between the nmes obtained by both approaches . we have calculated the nmes for the @xmath0 decay of the ca , ti and cr isotopic chains . the gt parts of the nmes are compared in fig . [ full_nme ] for sm and edf calculations . as in the case of actual @xmath0 decay candidates @xcite , the sm nmes are about half of the edf values . moreover , this difference is independent on the particular interaction used . for the sm , results with two effective interactions , kb3 g and gxpf1a , are shown , with differences of around @xmath37 . this agrees with and extends previous studies in the pf shell restricted to @xmath2ca @xcite . for the edf , we have also calculated nmes with the gogny d1 m functional , which results in very small differences with respect to the gogny d1s . figure [ full_nme ] also reveals that the trends along the isotopic chain are similar in sm and edf calculations . in particular , relative maxima are found in the decays of mirror nuclei : @xmath38ca@xmath39ti , @xmath40ti@xmath41cr and @xmath42cr@xmath43fe , in agreement with refs . maxima are more marked in sm calculations , where the initial and final states share the same isospin quantum number , @xmath44 . in the sm case the two states are exactly isospin - symmetric , because coulomb and isospin - symmetry - breaking terms in the nuclear interaction are neglected , but the overlap between mirror initial and final states is also maximal in the edf approach , which includes the coulomb term . for edf calculations , however , @xmath44 is not a good quantum number . the configuration space and nuclear correlations included in sm and edf calculations are very different , with the sm being able to take into account more general correlations but in a rather limited valence space . regarding the size of the configuration space it is important to note that in the pf shell the sm includes all orbitals with their corresponding spin - orbit partner . this is relevant because in the @xmath0 decay of heavier nuclei , some spin - orbit partners are not included in sm calculations , and this has been pointed out as a possible cause of the relatively small sm nmes . the sm calculations analyzed in this work are thus free from this shortcoming . we can get more insight in the comparison of sm and edf nmes by simplifying the nuclear structure correlations present in the initial and final states of the @xmath0 decay . figure [ sph_nme ] shows @xmath23 calculated with the same transition operator as fig . [ full_nme ] , but with simplified nuclear states . for the edf , spherical symmetry is assumed . in the sm case , only configurations with zero seniority ( @xmath45 ) are permitted , this is , protons and neutrons are coupled in @xmath19 pairs -no proton - neutron @xmath19 pairs are included . we observe that the gt parts of the nmes calculated in these simplified schemes are significantly larger than in the full calculation for both approaches , with an striking agreement between sm and edf nmes . indeed sm gxpf1a calculations lie within @xmath46 from edf values , while sm kb3 g calculations are about @xmath47 larger . the difference between the two sm results stems from the different @xmath19 , @xmath48 pairing . as shown in fig . [ full_nme ] , this difference between effective interactions is washed out when full calculations are performed . the agreement between sm and edf nmes is in strong contrast with the full nme calculations shown in fig . [ full_nme ] , where sm nmes were half of the edf values . this implies that the spherical edf and seniority - zero sm calculations , while conceptually very different , capture approximately the same physics , leaving out the nuclear structure correlations that reduce the @xmath0 decay nmes . these have been identified in refs . @xcite as the correlations associated with high - seniority components in the sm , and collective deformation effects in edf calculations . figure [ sph_nme ] also shows that the trends followed by the nmes calculated in both approaches are very similar , and indeed they follow to a good approximation the generalized seniority scheme in a single shell @xcite : @xmath49 where @xmath50 is the number of proton ( neutron ) pairs in the shell , @xmath51 the pair degeneracy and @xmath52 coefficients characteristic of a major shell . deviations from eq . ( [ gs ] ) are due to non - perfect shell closures and the @xmath33 dependence in the neutrino potentials . the `` inverted parabola '' from initial number of neutrons @xmath53 to @xmath54 , common to all cases , shows the filling of the neutron @xmath14 orbital associated to the shell closure at neutron number @xmath55 . the rather `` flat '' behavior between @xmath54 to @xmath56 is governed by the filling of the @xmath15 orbital , associated with the closure at @xmath57 . at this point , nmes obtained with the sm gxpf1a interaction decrease , due to the filling of the @xmath16 orbital associated with the @xmath58 closure , while the sm kb3 g and edf results , which do not predict such a closure , increase . in all cases the nmes at @xmath59 are larger because the @xmath17 orbital is starting to get filled . furthermore , eq . ( [ gs ] ) predicts that due to the filling of the proton @xmath14 shell ( @xmath60=4 ) , the nmes for ti and cr decays ( @xmath61 ) to be the same , as observed in fig . [ sph_nme ] . the fact that both seniority - zero sm and spherical edf calculations agree with the generalized seniority scheme , and result in very similar nmes , opens up the door to benchmarking also nmes calculated with other nuclear structure methods , such as qrpa or ibm , which using similarly simplified initial and final states should also agree with the results of fig . [ sph_nme ] . and @xmath23 , for ca@xmath3ti ( a)-(b ) , ti@xmath3cr ( c)-(d ) and cr@xmath3fe ( e)-(f ) @xmath0 decays . calculations are performed -using kb3 g interaction- with initial and final states obtained at zero - seniority ( red diamonds ) , zero - seniority with exact isospin projection ( blue inverted triangles ) and in the full calculation ( black left triangles ) . ] the role of nuclear structure correlations in @xmath23 is studied in figs . [ fig3_edf]- [ fig3_sm ] , where the full edf and sm results of fig . [ full_nme ] are compared to the spherical edf and seniority - zero sm results of fig . [ sph_nme ] . in addition , figs . [ fig3_edf]- [ fig3_sm ] also show intermediate results that give additional information on the role of correlations into @xmath0 decay nmes . within the edf approach we can explore the intrinsic quadrupole deformation @xmath62 in the initial and final nuclei , as well as the effect of shape mixing . [ fig3_edf](a ) shows potential energy surfaces ( pes , thin lines ) projected to particle - number and angular momentum for @xmath63ti@xmath64cr , and the ground - state collective initial and final states obtained after configuration mixing ( thick lines ) . [ fig3_edf](a ) distinguishes the three edf calculations in fig . [ fig3_edf ] , panels ( b ) , ( c ) and ( d ) . in the spherical calculation ( edf@xmath65 in fig . [ fig3_edf ] ) the initial and final states are the spherical @xmath66 states denoted with triangles in fig . [ . a better approach consists in considering the minima of the corresponding pes to calculate the nmes ( edf@xmath67 ) . finally , the full edf calculation uses self - consistent shape mixing of the collective states , within the gcm framework , to obtain the nmes ( edf@xmath68 ) . [ fig3_edf ] shows that the @xmath23 pattern found with edf spherical states disappears when pes minima are used . moreover , the nmes are significantly reduced when the deformation effects are included . furthermore , the full edf nmes roughly follow the trends of the pes minima solution , and configuration ( shape ) mixing only produces a shift to larger values , which is larger in the ti and cr decays after the neutron @xmath14 orbital is filled . on the other hand the sm calculations in fig . [ fig3_sm ] compare nmes obtained using the kb3 g interaction with seniority - zero initial and final states , the isospin projection of these states , and the full pf calculation . [ fig3_sm ] shows that isospin projection reduces significantly the fermi component of the nme -panels ( a ) , ( c ) and ( e)- but is only a small correction to @xmath23 -panels ( b ) , ( d ) and ( f ) . for the gt component , the reduction is maximal at @xmath69 nuclei , but very minor in the most neutron - rich systems . the correlations associated to high - seniority components in the initial and final states are the responsible for the strong reduction of @xmath23 , and these correlations also wash out the trend which appears with seniority - zero initial and final states . in addition , it follows figs . [ fig3_edf]- [ fig3_sm ] that these correlations reduce the nmes more significantly than the ones associated to collective deformation in the edf approach . figure [ isospin_nme ] gives a detailed account of the evolution of the sm @xmath23 and @xmath24 parts of the nmes as a function of the maximum seniority allowed in the initial and final nuclear states . this figure shows that for the @xmath42ca@xmath43ti @xmath0 decay , which relates two semi - magic nuclei , seniority components up to @xmath70 are necessary for a reliable @xmath23 and @xmath24 calculation . the seniority decomposition of the full sm states is 97%/3%/0% ( 77%/21%/2% ) for the @xmath45/@xmath70/@xmath71 components of @xmath42ca ( @xmath42ti ) . on the other hand , higher seniority components up to @xmath72 are needed in the @xmath2ti@xmath73cr decay . in this case the decomposition in seniority is 58%/37%/5% ( 27%/42%/31% ) for the @xmath45/@xmath70/@xmath71 parts in @xmath2ti ( @xmath2cr ) . high - seniority components are therefore associated with the description of the deformed @xmath2cr . spherical and full edf results are also shown in fig . [ isospin_nme ] . we have discussed above that spherical edf results roughly correspond to seniority - zero sm calculations . however , the full edf nmes behave quite differently in the two decays shown in in fig . [ isospin_nme ] . for @xmath42ca@xmath43ti decay , the final edf number agree with the results of the spherical nme calculation . this is due to the semi - magic character of the initial and final states , which prevents any collective correlation in these nuclei ( this also applies to the @xmath38ca@xmath39ti decay ) . on the contrary , the full nmes for the @xmath2ti@xmath73cr decay get contributions from collective deformation and shape mixing . these final nmes are roughly equivalent to the sm @xmath74 results . this suggests that correlations associated to high - seniority components in the sm are not completely captured in edf calculations . these could be partially responsible for the differences between sm and edf nmes shown in fig . [ full_nme ] . , panels a , b ) and fermi ( @xmath24 , panels c , d ) parts of the nuclear matrix element of the non - physical @xmath0 decays of @xmath42ca@xmath43ti ( panels a , c ) and @xmath2ti@xmath73cr ( panels b , d ) . shell model ( sm ) results are shown as a function of the maximum seniority permitted in the initial and final states ( squares ) , and also after isospin projection(circles ) . energy density functional ( edf ) results using spherical initial and final states ( dashed lines ) and the full edf calculation ( dashed - dotted lines ) are also shown . the edf gogny d1s and sm kb3 g interactions are used . ] since the edf states are built as linear combinations of projected hartree - fock - bogoliubov - type states with different axial quadrupole deformations , these intrinsic states are fully paired in time - reversed single - particle orbits by definition . therefore , pair - breaking _ in the seniority scheme _ is obtained by deforming the system , but not by including explicitly quasiparticle excitations on top of each intrinsic state . a step further , beyond the scope of this work , would include on equal footing both pair - breaking mechanisms into the gcm framework , and study their influence in the nmes . figure [ isospin_nme ] also shows that , when the seniority truncated @xmath23 results are projected to good isospin , they are mildly reduced . on the contrary , isospin projection has an important effect for @xmath24 , where states projected to good isospin are crucial . indeed , if the @xmath75 dependence of the neutrino potentials is removed , the @xmath24 connecting states with different @xmath44 values vanishes , as in two - neutrino double - beta decays . in conclusion , only small changes in the gt part of the nme are expected from projecting edf states to good isospin . this also applies to other methods calculating nmes which break isospin symmetry , such as the qrpa and ibm . [ summary ] we have studied the gt part of the nmes of the ca@xmath3ti , ti@xmath3cr and cr@xmath3fe @xmath0 decays . the systematic study of these non - physical decays allows us to compare shell - model and energy density functional calculations . we observe that when full sm and edf calculations are performed , sm results are about half the edf values . however , when we simplify the initial and final states of the decay to spherical edf and seniority - zero sm states , the nmes obtained by both approaches are surprisingly similar , suggesting that the nuclear structure description is equivalent for both methods at this level . we have studied the role of nuclear structure correlations to the nmes , and we note that in general correlations associated to high - seniority sm components and edf collective deformation reduce the nmes . a comparison between these two suggests that the correlations associated to higher - seniority sm components are not completely captured by the edf approach , pointing to a possible reason for the difference between sm and edf nmes . we have also explored projection to good isospin of the initial and final @xmath0 decay states , and conclude that , unlike for the fermi part , it has only a moderate effect in the gamow - teller part of the nmes . this work opens up the door for benchmarks between nme calculated within different theoretical approaches , and constitutes a step forward towards identifying the relevant ingredients that will lead to reliable nme calculations with reduced theoretical uncertainties . this work was partly supported by the helmholtz association through the helmholtz alliance program , contract ha216/emmi `` extremes of density and temperature : cosmic matter in the laboratory '' , the helmholtz international center for fair within the framework of the loewe program launched by the state of hesse , by the deutsche forschungsgemeinschaft through contract sfb 634 and by the bmbf - verbundforschungsprojekt number 06da7047i . ap is partially supported by the micinn ( spain ) ( fpa2011 - 29854 ) ; by the comunidad de madrid ( spain ) ( hephacos s2009-esp-1473 ) and by the european union fp7 itn invisibles ( marie curie actions , pitn- ga-2011- 289442 ) . 9 f. t. avignone , s. r. elliot , j. engel , rev . mod . phys . 80 , 481 ( 2008 ) . ( exo collaboration ) , phys . 109 , 032505 ( 2012 ) . ( kamland - zen collaboration ) , phys . 110 , 062502 ( 2013 ) . m. agostini _ ( gerda collaboration ) , phys . 111 , 122503 ( 2013 ) . k .- h . et al . _ , j. c 73 , 2330 ( 2013 ) . d. g. phillips ii _ et al . _ , j. phys . ser . 381 012044 ( 2012 ) . et al . _ , j. phys . : ser . 375 042018 ( 2012 ) . k. zuber , prog . part .nucl . 64 , 267 ( 2010 ) . et al . _ , aip conf . 942 , 101 ( 2007 ) . j. argyriades _ et al . c 80 , 032501(r ) ( 2009 ) . _ , j. phys . 375 , 042023 ( 2012 ) . c. arnaboldi _ et al . _ , c 78 , 035502 ( 2008 ) . v. lvarez _ et al . _ , jinst 8 , p04002 ( 2013 ) . j. kotila and f. iachello , phys . c 85 , 034316 ( 2012 ) . s. stoica and m. mirea , phys . c 88 , 037303 ( 2013 ) . e. caurier , j. menndez , f. nowacki , and a. poves , phys . 100 , 052503 ( 2008 ) . j. menndez , e. caurier , f. nowacki , and a. poves , nucl . phys . a 818 , 139 ( 2009 ) . t. r. rodrguez and g. martinez - pinedo , phys . 105 , 252503 ( 2010 ) . n. lpez - vaquero , t. r. rodrguez and l. egido , phys . 111 , 142501 ( 2013 ) . f. simkovic , v. rodin , a. faessler , and p. vogel , phys . c 87 , 045501 ( 2013 ) . j. suhonen and o. civitarese , j. phys . g : nucl . part . 39 185105 ( 2012 ) . j. barea , j. kotila and f. iachello , phys . c * 87 * , 014315 ( 2013 ) . p. vogel , j. phys . g : nucl . part . 39 124002 ( 2012 ) . s. j. freeman and j. p. schiffer , j. phys . g : nucl . part . 39 124004 ( 2012 ) . b. p. kay _ et al . _ , c 87 , 011302(r ) ( 2013 ) . d. frekers , p. puppe , j. h. thies and h. ejiri , nucl . a 916 , 219 ( 2013 ) . e. caurier , f. nowacki , and a. poves , phys . lett b 711 , 62 ( 2012 ) . f. simkovic , a. faessler , and p. vogel , phys . c 79 , 015502 ( 2009 ) . j. menndez , e. caurier , f. nowacki , and a. poves , phys . c 80 , 048501 ( 2009 ) . j. suhonen and o. civitarese , nucl . a 847 , 207 ( 2010 ) t. r. rodrguez and g. martinez - pinedo , phys . lett . b 719 , 174 ( 2013 ) . e. caurier , g. martnez - pinedo , f. nowacki , a. poves , and a. p. zuker , rev . phys . * 77 * , 427 ( 2005 ) . f. simkovic _ et al . _ , c 60 , 055502 ( 1999 ) . p. ring , p. schuck , _ the nuclear many body problem _ , springer - verlag , berlin , 1980 . j. f. berger et _ al . a * 428 * , 23 ( 1984 ) . delaroche et _ al . c 81 , 014303 ( 2010 ) . a. poves , j. snchez - solano , e. caurier , and f. nowacki , nucl . * a694 * , 157 ( 2001 ) . m. t. honma , t. otsuka , b. a. brown , and t. mizusaki , phys . c * 69 * , 034335 ( 2004 ) . j. d. holt and j. engel , phys . c * 87 * , 064315 ( 2013 ) . a. a. kwiatkowski _ et al . _ , arxiv 1308.3815 . m. kortelainen , j. suhonen , phys . rev c 75 , 051303(r ) ( 2007 ) . f. simkovic , r. hodk , a. faessler , and p. vogel , phys . c 83 , 015502 ( 2011 ) . r. a. senkov and m. horoi , arxiv:1310.3524 . h. feldmeier , t. neff , r. roth , j. schnack , nucl . phys . a * 632 * , 61 ( 1998 ) . m. kortelainen , o. civitarese , j. suhonen , and j. toivanen , phys . lett b 647 , 128 ( 2007 ) . f. simkovic _ c 79 , 055501 ( 2009 ) . j. menndez , d. gazit , and a. schwenk , phys . 107 , 062501 ( 2011 ) . e. caurier , f. nowacki , and a. poves , eur . phys . j. a 36 195 ( 2008 ) . m. horoi and s. stoica , phys . c 81 , 024321 ( 2010 ) . j. menndez , a. poves , e. caurier , and f. nowacki , j. phys . ser . * 267 * 012058 ( 2011 ) . j. barea and f. iachello , phys . c * 79 * , 044301 ( 2009 ) .	we calculate the nuclear matrix elements ( nmes ) for neutrinoless double - beta decays ( @xmath0 ) of @xmath1-shell nuclei using the shell model ( sm ) and energy density functional ( edf ) methods . the systematic study of non - physical decays ( except for @xmath2ca ) of ca@xmath3ti , ti@xmath3cr and cr@xmath3fe allows for a detailed comparison between the two nuclear structure approaches . we observe that while the dominant gamow - teller part of the nme differs roughly by a factor of two between sm and edf , when we restrict the calculations to spherical edf states and seniority - zero sm configurations , the nmes obtained by both methods are strikingly similar . this points out to the important role of nuclear structure correlations for @xmath0 decay nmes . we identify correlations associated with high - seniority components in the initial and final states of the decay as one of the reasons for the discrepancies between sm and edf results . we also explore exact projection to good isospin , and conclude that it has only a moderate effect in the gamow - teller part of the nmes but strongly affects the fermi contribution . this work opens up the door for nme benchmarks between different theoretical approaches , and constitutes a step forward towards more reliable estimations of the nmes .
the octahedron recurrence is the product of three chains of research seeking a common generalization . the first is the study of the algebraic relations between the connected minors of a matrix , and particularly of a recurrence relating them known and dodgson condensation . attempting to understand the combinatorics of dodgson condensation lead to the discovery of alternating sign matrices and aztec diamonds . aztec diamonds are graphs whose perfect matchings have extremely structured combinatorics and soon formed their own , second line of research as other graphs were discovered with the similar regularities . the third is the study of somos - sequences and the laurent phenomenon , which began with an attempt to understand theta functions from a combinatorial perspective . in this section we will sketch both lines of research . in the following section , we will begin to describe the vocabulary and main results of this paper . let @xmath0 be a collection of formal variables indexed by ordered pairs @xmath1 . let @xmath2 , @xmath3 , @xmath4 and @xmath5 be integers with @xmath6 and set @xmath7 charles dodgson @xcite observed that @xmath8 suggested using this recursion to compute determinants and found generalizations that could be used when the above recursion requires dividing by @xmath9 . ( this formula is not original to dodgson , according to page 7 of @xcite this formula has been variously attributed to sylvester , jacobi , desnanot , dodgson and frobenius . however , the use of this identity iteratively to compute connected minors and the detailed study of this method appears to have begun with dodgson and is known as dodgson condensation . ) before passing on , we should note that the study of algebraic relations between determinants and how they are effected by various vanishing conditions is essentially the study of the flag manifold and schubert varieties . this study lead to the invention of cluster algebras which later returned in the laurentness proofs we will discuss in section [ somos ] . see @xcite for the definitions of cluster algebras . the precise relation between cluster algebras and ( double ) schubert varieties is discussed in @xcite , although the reader may wish to first consult the references in that paper for the history of results that preceeded it . one can visualize this recurrence as taking place on a three dimensional lattice . the values @xmath0 are written on a horizontal two dimensional plane . the values @xmath10 live on a larger three dimensional lattice with @xmath10 written above the center of the matrix of which @xmath10 is the determinant and at a height proportional to @xmath11 . the six entries involved in computing the recurrence lie at the vertices of an octahedron . we make a change of the indexing variables of the lattice so that the lattice consists of those @xmath12 with @xmath13 . then the recurrence above can be rewritten as @xmath14 with @xmath15 and @xmath16 . robbins and rumsey @xcite considered modifying the recurrence to @xmath17 and also generalized by allowing @xmath18 to be arbitrary instead of being forced to be @xmath19 . we write @xmath20 when @xmath21 is even and @xmath22 when @xmath21 is odd . robbins and rumsey discovered and proved that , despite these modifications , the functions @xmath23 were still laurent polynomials in the @xmath24 and in @xmath25 , with all coefficients equal to @xmath19 . robbins and rumsey discovered that the exponents of the @xmath24 in any monomial occuring in a @xmath26 formed pairs of `` compatible alternating sign matrices . '' we will omit the definition of these obects in favor of discussing a simpler combinatorial description , found by elkies , kuperberg , larsen and propp @xcite , in terms of tilings of aztec diamonds . if @xmath27 is a graph then a _ matching _ of @xmath27 is defined to be a collection @xmath28 of edges of @xmath27 such that each vertex of @xmath27 lies on exactly one edge of @xmath28 . matchings are also sometimes known as `` perfect matchings '' , `` dimer covers '' and `` 1-factors '' . fix integers @xmath29 , @xmath30 and @xmath31 with @xmath32 . in this section , we will describe a combinatorial interpretation of the entries of @xmath26 in terms of matchings of certain graphs called aztec diamonds . for simplicity , we take @xmath33 in the recurrence for @xmath34 . consider the infinite square grid graph : its vertices are at the points of the form @xmath35 with @xmath1 , its edges join those vertices that differ by @xmath36 or @xmath37 and its faces are unit squares centered at each @xmath1 . we will refer to the face centered at @xmath38 as @xmath38 . consider the set @xmath39 of all faces @xmath38 with @xmath40 . let @xmath41 be the induced subgraph of the square grid graph whose vertices are adjacent to some face of @xmath39 . we will call this the aztec diamond of order @xmath29 centered at @xmath42 . we define the _ faces _ of @xmath41 to be the squares @xmath38 for @xmath43 . note that this includes more squares than those in @xmath39 ; in particular , it contains some squares only part of whose boundary lies in @xmath41 . we refer to the faces in @xmath39 as _ closed _ faces and the others as _ open _ faces . let @xmath28 be a matching of @xmath41 . if @xmath38 is a closed face then either @xmath9 , @xmath19 or @xmath44 edges of @xmath38 are used in @xmath28 ; we define @xmath45 to be @xmath46 , @xmath9 or @xmath19 respectively . if @xmath24 is an open face then either @xmath9 or @xmath19 edges will be used and , we define @xmath45 to be @xmath9 or @xmath19 respectively . set @xmath47 , where the product is over all faces @xmath38 of @xmath41 . @xmath48 , where the sum is over all matchings @xmath28 of @xmath41 . it seems to be difficult to find a reference which states this result exactly in this form . @xcite shows that @xmath26 is equal to a sum over `` compatible pairs of alternating sign matrices '' and @xcite shows that such pairs are in bijection with perfect matchings of the aztec diamond . tracing through these bijections easily gives the claimed theorem . another proof can be found by mimicing the method used in @xcite to count the matchings of the aztec diamond . this result will also be a corollary of this paper s main result . the order @xmath49 aztec diamond has @xmath50 perfect matchings . clearly , if we plug in @xmath19 for each @xmath24 , then @xmath51 for every matching @xmath28 so the number of matchings of the order @xmath49 aztec diamond will be @xmath52 . we will denote this number by @xmath53 . then the octahedron recurrence gives @xmath54 and the claim follows by induction . many other families of graphs have been discovered for which the number of matchings of the @xmath55 graph is proportional to a constant raised to a quadratic in @xmath49 . they are summarized and unified in @xcite . one of the purposes of this paper is to give proofs of these formulae as simple as the corrolary above . the monomial @xmath56 keeps track of the matching @xmath28 in a rather cryptic manner ; it is not even clear that @xmath56 determines @xmath28 , although this will follow from a later result ( proposition [ facedetermine ] ) . it is natural to add new variables that code directly for the presence or absence of individual edges . let @xmath57 and @xmath58 be integers such that @xmath59 . draw the plane such that @xmath57 increases in the east - ward direction and @xmath58 in the north - ward . let the edges east , north , west and south of @xmath38 be labelled @xmath60 , @xmath61 , @xmath62 and @xmath63 respectively . introduce new free variables with the same names as the edges . we now define an enhanced version of @xmath34 by @xmath64 where the sum is still over all matchings of the order @xmath29 aztec diamond . each matching is now counted by both the monomial @xmath56 and the product of the edges occuring in that matching . the new @xmath34 obeys the recurrence @xmath65 somos has attempted to rebuild the theory of theta functions on a combinatorial basis . theta functions are functions of two variables , traditionally called @xmath66 and @xmath67 . the simplest theta function is @xmath68 . from now on , we will concentrate only on the dependence on @xmath66 and so write @xmath69 . for a general introduction to @xmath70 functions , see for example chapter xxi of @xcite . somos began with a well known result , that , for any @xmath71 and @xmath72 , @xmath69 would obey @xmath73 for @xmath74 and @xmath75 certain constants depending in a complicated manner on @xmath67 . somos proposed using this recurrence to compute @xmath76 in terms of @xmath74 , @xmath75 and @xmath77 , @xmath78 , @xmath79 and @xmath80 . he discovered that @xmath76 was a laurent polynomial in these @xmath81 variables . this is not difficult to prove by elementary means , but suggests a deeper explanation may be lurking . moreover , he discovered that all of the coefficients of this laurent polynomial were positive but was unable to prove this ; this will be proven for the first time as a corollary of the results of this paper . after somos s work , many other recurrence with surprising laurentness properties were discovered ; see @xcite for a summary . the family of most importance for this paper is the three term gale - robinson recurrence : let @xmath82 , @xmath83 and @xmath84 be positive integers with @xmath85 . the three term gale - robinson recurrence ( abbreviated to `` gale - robinson recurrence '' in the remainder of this paper ) is @xmath86 note that the somos recurrence is the case @xmath87 . note that if @xmath88 and @xmath89 , we get the recurrence @xmath90 , the recurrence for the number of perfect matchings of the aztec diamond . the gale - robinson recurrence , for every value of @xmath91 , can be thought of as a special case of the octahedron recurrence . this observation was first made by propp and seems to have first appeared in print in @xcite . the reduction is performed as follows : suppose that @xmath92 obeys the gale - robinson recurrence . define @xmath23 by @xmath93 . then @xmath34 obeys @xmath94 thus , if we choose @xmath60 , @xmath61 , @xmath62 and @xmath63 to be constants @xmath83 , @xmath84 , @xmath95 and @xmath96 with @xmath97 and @xmath98 , @xmath34 obeys the octahedron recurrence . in these circumstances , we must study the octahedron recurrence not with the initial conditions @xmath99 and @xmath18 but with the initial coniditions @xmath23 for @xmath12 such that @xmath100 . it therefore beomes our goal to study the octahedron recurrence with general initial conditions . the main result of this paper will be that , with any inital conditions , @xmath23 is a laurent polynomial in the initial coniditions and in the coefficients @xmath60 , @xmath61 , @xmath62 and @xmath63 . moreover , all of the coefficients of this laurent polynomial are @xmath19 . in particular , they are positive . in the reduction of the gale - robinson recurrence to the octahedron recurrence , we imposed that many of the initial conditions of the octahedron recurrence be equal to each other . thus , many monomials that would be distinct if the octahedron recurrence were run with arbitrary inital conditions become the same in the gale - robinson recurrence . however , we can still deduce that all of the coefficients of the laurent polynomials computed by the gale - robinson recurrence are positive . in this section we will introduce the main objects and results of this paper . the main object of study of this paper is a function @xmath34 defined on a three dimensional lattice by a certain recurrence . we denote the set of initial conditions from which @xmath34 is generated by @xmath101 and consider varying the shape of @xmath101 inside this lattice . it turns out that the values of @xmath34 are laurent polynomials in the initial variables where every term has coefficient 1 . we are able further more to find families of graphs such that @xmath34 gives generating functions for the perfect matchings of these graphs . special cases of this result include being able to choose @xmath101 so that @xmath34 gives us the values of three term gale - robinson sequences or such that the families of graphs are aztec diamonds , fortresses and other common families of graphs with simple formulas for the number of their matchings . in the next subsection , we introduce the vocabulary necessary to define @xmath34 . in the following subsection , we introduce the vocabulary necessary to define the families of graphs . set @xmath102 and set @xmath103 where @xmath67 denotes one of the four symbols @xmath83 , @xmath84 , @xmath95 and @xmath96 . @xmath104 , @xmath105 and @xmath106 stand for `` lattice '' , `` edges '' and `` faces . '' we call @xmath107 and @xmath108 `` lattice - adjacent '' if @xmath109 . ( another notion of adjacency will arise later . ) for @xmath110 , set @xmath111 ( we will often drop the subscript on @xmath112 when it is clear from context . ) let @xmath113 we call @xmath114 , @xmath115 and @xmath116 the _ cone _ , _ inner cone _ and _ outer cone _ of @xmath117 respectively . @xmath114 is a square pyramid with it s vertex at @xmath117 . let @xmath118 and define @xmath119 assume that 1 . @xmath120 if @xmath121 and @xmath122 are lattice - adjacent . 2 . @xmath123 3 . @xmath124 . the last condition is equivalent to `` for any @xmath125 , @xmath126 is finite . '' we will call such an @xmath127 a `` height function '' and call a function @xmath127 that obeys the first two conditions a pseudo - height function . note : this definition is not related to the height functions in the theory of aztec diamonds , described in @xcite . let @xmath128 be the field of formal rational functions in the following infinite families of variables : the family @xmath24 , @xmath57 , @xmath129 and the families @xmath60 , @xmath61 , @xmath62 , @xmath63 where @xmath57 , @xmath129 and @xmath130 . ( clearly , @xmath106 indexes the @xmath131 s and @xmath105 indexes the @xmath83 s , @xmath84 s , @xmath95 s and @xmath96 s ) . let @xmath132 be the ring @xmath133.$ ] for reasons to appear later , we call the @xmath131 s the `` face variables '' and the @xmath83 s , @xmath84 s , @xmath95 s and @xmath96 s the `` edge variables . '' we define a function @xmath134 recursively by @xmath135 for @xmath136 and @xmath137 for @xmath138 . our third condition on @xmath127 ensures that this recurrence will terminate . the main result of this paper is @xmath139 and , when written as a laurent polynomial , each term appears with coefficient 1 . moreover , the exponent of each face variable is between @xmath46 and @xmath140 ( inclusive ) and the exponent of each edge variable is @xmath9 or @xmath19 . to describe our result more precisely , we need to introduce some graph - theoretic terminology . let @xmath141 be a connected planar graph with a specified embedding in the plane and let @xmath142 , @xmath143 , @xmath144 be the edges of the outer face in cyclic order . we define a `` graph with open faces '' to be such a graph @xmath141 with a given partition of the cycle @xmath142 , @xmath144 into edge disjoint paths @xmath145 , @xmath146 , @xmath147 . denote by @xmath27 the graph with open faces associated to @xmath141 and the partition above . we call a path @xmath145 , @xmath147 in the given partition an `` open face '' of @xmath27 . we denote the open faces of @xmath27 by @xmath148 . we refer an interior face of @xmath141 as a `` closed face '' of @xmath27 and denote the set of them by @xmath149 . we set @xmath150 and call a member of @xmath151 a face of @xmath27 . note that the exterior face of @xmath141 is _ not _ usually a face of @xmath27 . the image associated to a is that open disks have had portions of their boundary glued to the outside of @xmath141 along the paths @xmath145 , @xmath147 with some additional boundary left hanging off into space . we will at times need to allow @xmath141 to be infinite , in which case @xmath141 might have several outer faces , or none at all . we will not describe the appropriate modifications , as they should be obvious . the infinite graphs which will come up in this paper will have no outer face and the corresponding will have no open faces . we will refer to an edge or vertex of @xmath141 as an edge or vertex ( respectively ) of @xmath27 . we say that an edge @xmath152 borders an open face @xmath34 of @xmath27 if @xmath152 lies in the associated path @xmath145 , @xmath147 . all other incidence terminology ( endpoint , adjacency of faces , etc . ) should be intuitive by analogy to ordinary planar graphs . we use @xmath153 and @xmath154 to denote the edges and vertices of @xmath27 . by a map @xmath155 we mean a triple of injections @xmath156 , @xmath157 and @xmath158 compatible with the adjacency relations . we say @xmath159 is a sub - if such a map exists . note that we do not require that open faces be taken to open faces . let @xmath160 be the ring @xmath161 $ ] where @xmath162 are formal variables indexed by @xmath163 and @xmath164 are formal variables indexed by @xmath165 . for the rest of this subsection , assume that @xmath27 is finite . a matching of @xmath27 is a collection of edges @xmath28 such that every vertex lies on exactly one edge of @xmath28 . let @xmath28 be a matching of @xmath27 . for @xmath166 , set @xmath167 if @xmath168 , @xmath9 otherwise . for @xmath163 , let @xmath83 be the number of edges of @xmath34 that lie @xmath28 and @xmath84 the number of edges of @xmath34 not in @xmath28 . if @xmath169 , set @xmath170 if @xmath171 set @xmath172 ( note that @xmath173 . ) for any matching @xmath28 , set @xmath174 for any @xmath27 , set @xmath175 where the sum is over all matchings of @xmath27 . we call @xmath56 the matching monomial of @xmath28 and @xmath176 the matching polynomial of @xmath27 . it will turn out that the actual graphs to which we will apply this definition are bipartite , so every closed face has an even number of edges and we may omit the @xmath177 in this case . however , it will be convenient to have a definition that is valid for all graphs . we can now give our second statement of the main theorem . for any height function @xmath127 we can find an infinite @xmath178 , a decomposition @xmath179 and injective maps @xmath180 , @xmath181 and a family of sub-@xmath182 indexed by @xmath138 such that 1 . if we define @xmath183 by @xmath184 for @xmath185 , @xmath186 where @xmath187 for @xmath188 and @xmath189 for @xmath190 , then @xmath191 ( we have implicitly used the obvious injection @xmath192 . ) 2 . if @xmath193 then @xmath194 is a sub - of @xmath182 we refer to the edges in @xmath195 as the `` weighted edges '' and the edges in @xmath196 as the `` unweighted edges '' . clearly , this will imply the previous statement of the main theorem , except for the bounds on the exponents and the claim that the coefficients are 1 . the first will follow by showing as well that every face of @xmath27 has @xmath197 edges . the second will follow by showing that the edges of @xmath196 are vertex disjoint , so a matching @xmath28 is uniquely determined by @xmath198 . in section [ algorithm ] , we will describe an algorithm we refer to as `` the method of crosses and wrenches '' for finding the graph @xmath178 and the subgraphs @xmath199 . we will postpone proving the correctness of the algorithm to present applications of the theory in section [ examples ] . in particular , we will carry out the examples of somos-4 and somos-5 and show explicitly the families of graphs associated to them . we will also show that , by choosing certain periodic functions for @xmath127 , we can create fortress graphs and families of graphs studied by chris douglass and matt blum . the number of matchings of these graphs are powers of 5 , 2 and 3 respectively , in each case we will give a rapid proof of this by induction from our main theorem . fortresses and douglass graphs are discussed in @xcite and these formulas are also proved there . in the section [ urban ] , we will first relate the matchings of the graphs @xmath199 to the matchings of certain infinite graphs subject to `` boundary conditions at infinity '' . this will remove the elegant property that @xmath200 when @xmath201 , but it will create objects better suited to an inductive proof . we will then prove the main theorem by varying @xmath101 and holding @xmath12 fixed . in the process , we will recover variants of the `` urban renewal''operations of @xcite . in section [ kuo ] , we will give a second proof , inspired by a proof of @xcite for the aztec diamond case , that works by holding @xmath101 fixed and varying @xmath12 . this proof has some additional consequences that the first proof does not . however , due to the extraordinary number of cases that would otherwise be involved , we will not check that all of the exponents work out . in section [ final ] we make some final comments . one might wonder whether the introduction of so many families of variables is necessary to the paper . for most parts of the paper , one may think of any family variables one does not want to deal with as simply being set equal to @xmath19 without difficulty . however , there is one place where this does not work : our first proof of the main theorem is inductive and the induction only goes through if the theorem is stated with the entire family of face variables present . most of this research was done while i was working in reach ( research experiences in algebraic combinatorics at harvard ) and was aided by many helpful conversations with the other members of reach , and in particular by daniel abramson , gabriel carroll and seth kleinerman s willingness to work through many special cases with me . above all , i would like to thank jim propp for suggesting this problem and showing me much of his unpublished work on it . in this section we describe how to find the graphs @xmath178 and @xmath199 discussed above . as a running example , we we ll compute @xmath202 with @xmath203 . the goal is to predict the formula : @xmath204 the four terms in the above formula will eventually be shown to correspond to the four matchings of the shown in figure [ example ] . here the dashed edges seperate the various open faces and the graph is drawn twice so that the face and edge labels will not overlap each other . the four matchings are shown in figure [ matchings ] ; the numbers in figure [ matchings ] are the exponents of the corresponding face variables . the octahedron recurrence with these initial conditions appears section 9.4 of @xcite . a `` tropicalized '' version of our running example appears in @xcite . let @xmath205 be a height function . we describe the graph @xmath178 by describing its dual . the faces of @xmath178 ( which are all closed ) are indexed by the elements of @xmath101 and the map @xmath206 sends the face @xmath136 to @xmath207 . picture the face @xmath12 as centered at the point @xmath208 . if @xmath107 and @xmath122 are lattice - adjacent ( i.e. @xmath209 ) then @xmath210 borders @xmath211 . in addition , if @xmath212 then @xmath213 borders @xmath211 iff @xmath214 . no other pairs of faces border and faces which border border only along a single edge . we refer to this method of finding @xmath178 as the `` method of crosses and wrenches '' because it can be describe geometrically by the following procedure : any quadruple of values @xmath215 must be of one of the following six types : @xmath216 in the center of these squares , we draw a , or in the first and second , third and fourth , and fifth and sixth cases respectively . the diagram below displays the possible cases . @xmath217 we then connect the four points protruding from these symbols by horizontal and vertical edges . ( we often have to bend the edges slightly to make this work . kinks introduced in this way are not meant to be vertices , all the vertices come from the center of a @xmath218 or from the two vertices at the center of a @xmath219 or @xmath220 . ) we refer to the @xmath218 s as `` crosses '' and the @xmath219 s and @xmath220 s as `` wrenches '' . in our running example , if @xmath221 then @xmath222 and we place a @xmath218 . otherwise , we get a @xmath219 . the resulting infinite graph , shown in figure [ infiniteexample ] , consists of a diagonal row of quadrilaterals seperating a plane of hexagons . we will now describe the decomposition @xmath223 and the map @xmath180 . the set @xmath195 will consist of the horizontal and vertical edges , i.e. , those separating lattice - adjacent faces . the set @xmath196 will consist of the diagonal edges , i.e. , those which come from the center of a wrench . consider any edge @xmath152 of @xmath195 . such an edge lies between two faces @xmath213 and @xmath211 with @xmath224 . without loss of generality , @xmath225 . ( otherwise , switch the names of @xmath107 and @xmath226 . ) there are four cases : 1 . if @xmath227 and @xmath228 then @xmath229 . 2 . if @xmath227 and @xmath230 then @xmath231 . 3 . if @xmath232 and @xmath228 then @xmath233 . 4 . if @xmath232 and @xmath230 then @xmath234 . the reader may wish to refer to figure [ example ] . finally , we must describe the sub-@xmath199 of @xmath178 that corresponds to a particular @xmath12 . we will abbreviate @xmath199 by @xmath27 in this paragraph . the closed faces of @xmath27 will be @xmath235 . the edges of @xmath27 will be the edges of @xmath178 adjacent to some face in @xmath236 . the open faces of @xmath27 will be those members of @xmath101 some but not all of whose edges are edges of @xmath27 . note that every open face of @xmath27 lies in @xmath237 but the converse does not necessarily hold . note also that there are never any edges that separate two open faces ; even if those two open faces are lattice - adjacent . another way to describe the faces of @xmath27 which is more intuitive but harder to write down is that the faces of @xmath27 correspond to those @xmath238 which are used in compuing @xmath23 and the closed faces correspond to those @xmath238 which are divided by in the course of this computation . in our running example , the closed faces are @xmath239 , @xmath240 and @xmath241 . the open faces are @xmath242 , @xmath243 , @xmath244 , @xmath245 , @xmath246 , @xmath247 and @xmath248 . we can now give the final statement of our main theorem . for any height function @xmath127 , define @xmath178 , @xmath206 and @xmath199 by the algorithm of the previous sections . then the second statement of the main theorem holds with regard to these choices . it is easy to check that every closed face of @xmath178 has 4 , 6 or 8 sides . ( simply check all possible values for the face s eight neighbors . ) hence all of the @xmath199 are bipartite . moreover , every face has @xmath197 edges , as promised above . it is also easy to fulfill another promise and check that the unweighted edges are vertex disjoint : they lie in the middle of wrenches and do not border each other . it is clear that @xmath249 is injective ( and in fact , bijective ) . we now show : the map @xmath180 is injective . consider an edge @xmath250 , the cases of @xmath84 , @xmath95 and @xmath96 are extremely similar . any edge with label @xmath152 must be between two faces of the form @xmath251 and @xmath252 . we must show there is at most one value of @xmath49 for which @xmath253 and @xmath252 both lie in @xmath101 . suppose , for contradiction , there are two such values : @xmath49 and @xmath254 . without loss of generality , suppose that @xmath255 . then @xmath256 but @xmath205 must change by @xmath257 when @xmath57 or @xmath58 changes by 1 , so @xmath127 can not change by this much between @xmath258 and @xmath259 , a contradiction . we now describe where the faces of @xmath260 lie in the plane . we make use of the following abbreviations and notations : we shorten @xmath260 to @xmath27 and @xmath261 to @xmath115 . recall the notation @xmath262 . so @xmath263 . let @xmath207 be a closed face of @xmath27 and let @xmath264 be such that @xmath265 is between @xmath57 and @xmath30 and @xmath266 is between @xmath58 and @xmath31 . then @xmath267 is also a closed face of @xmath27 . without loss of generality , we may assume that @xmath268 and @xmath269 . we have @xmath270 . on the other hand , we have @xmath271 . as @xmath272 , we have @xmath273 . so the faces of @xmath27 form four `` staircases '' , as in figure [ facelayout ] . as a corollary , we deduce @xmath27 is connected . clearly , if @xmath274 and @xmath275 lie on the same closed face of @xmath27 , then @xmath274 and @xmath275 lie in the same connected component of @xmath27 . but , by the previous proposition , we can travel from any closed face of @xmath27 to @xmath276 along a sequence of lattice - adjacent faces and lattice adjacent faces have a vertex in common . thus , since every vertex of @xmath27 lies on a closed face , every vertex must lie in the same connected component as the vertices of @xmath42 . so @xmath27 is connected . it is also clear from the picture above that @xmath27 is bounded by a single closed loop . call this loop @xmath277 . divide @xmath277 , as in figure [ facelayout ] , into four arcs by horizontal and vertical lines through @xmath276 . ( in the diagram , @xmath277 is in bold and the lines through @xmath42 are dashed . ) [ boundary ] these lines divide @xmath277 into four paths . each of these paths contains an odd number of vertices . on each of these paths , the vertices alternate between vertices all of whose neighbors in @xmath178 are inside or on @xmath277 ( and thus in @xmath27 ) and vertices all of whose neighbors in @xmath178 are outside or on @xmath277 the ends of each path are of the latter type . we consider the section of @xmath277 on which @xmath278 and @xmath279 , the other four sections are similar . for the faces @xmath38 of @xmath178 within @xmath277 ( that is the closed faces of @xmath27 ) we have @xmath280 where as for the faces which border @xmath277 on the outside ( that is , the open faces of @xmath27 ) we have @xmath281 . our proof is by induction on the number of faces @xmath38 of @xmath27 for @xmath282 and @xmath279 . in the base case , where there is only one such face , the path in question has one vertex and the result is trivial . when there is more then one such face , we can always find a face @xmath38 such that @xmath283 and @xmath284 are outside @xmath277 . for simplicity , we describe the situation where @xmath285 and @xmath286 and leave the boundary cases to the reader . set @xmath287 . we must have @xmath288 as they are open faces of @xmath27 . we must have @xmath289 ; as @xmath290 we must have @xmath291 . if we then change @xmath205 to @xmath127 , we remove @xmath38 from within @xmath277 . there are four cases , based on whether @xmath292 and @xmath293 or @xmath294 ; figure [ boundarychange ] shows what happens in each case . the edges of @xmath277 are drawn in bold , the faces of @xmath27 are surrounded by thin boxes and the bipartite coloring of @xmath27 is shown in black and white . we will use this proposition in the proof of proposition [ geomn ] . this proposition is also useful in testing whether it is likely that a graph can be realized through the method of crosses and wrenches . for example , the hexagonal graphs of section 6 of @xcite can not be so realized , because on their boundary there are six different boundary edges connecting vertices of degree two and a crosses and wrenches graph can only have four such edges . examples [ examples ] we begin with an example we discussed in the introduction : the aztec diamond graphs . in this case , @xmath295 or @xmath46 when @xmath296 is @xmath9 or @xmath19 respectively . in this case , every square of four values is @xmath297 so in every case we get a cross . @xmath298 the infinite graph @xmath178 is just the regular square grid . the edge labeling is exactly as described in section [ aztecintro ] . the graphs @xmath199 are the standard aztec diamond graphs . in this section we show that crosses and wrenches graphs are capable of reproducing several previously studied families of graphs ; the number of whose matchings are perfect powers or near perfect powers . inside each face @xmath38 , we put the number @xmath205 . most of these families of graphs appear to have first been described in print in @xcite , which we will use as our reference . fortress graphs are discussed in @xcite , section four . to obtain a fortress graph , we take @xmath299 every square of values is of the form @xmath300 so in every case we get a wrench , in the repeating pattern @xmath301 joining up the wrenches , we get the infinite graph in figure [ fortress ] . once again , if we set all of the variables equal to 1 , @xmath23 will count the matchings of fortresses . we must have @xmath302 and @xmath34 must obey the defining recurrence . it is easy to check that both conditions are obeyed by @xmath303 a result of @xcite . for example , figure [ fortexamp ] shows a fortress with 25 matchings , it is of the form @xmath304 where the center square is @xmath38 and @xmath305 , @xmath306 . this fortress has 9 closed and 16 open faces . similarly , to get another family of graphs first studied by chris douglas ( see @xcite , section six ) , we take @xmath307 we now sometimes will have @xmath308 which will produce a cross and sometimes will have @xmath309 which will produce a @xmath220 . overall , we have the repeating pattern @xmath310 which produces the infinite graph in figure [ douglass ] again , a quick induction verifies that @xmath311 similarly , to get the graphs considered by matthew blum ( see @xcite , section 8) , take @xmath312 ( ciucu s methods require him to embed @xmath178 into the square grid , and he thus takes @xmath178 to be made of octagons and hexagons . for our purposes , it is more natural to apply lemma [ split ] of section [ basiclemmas ] to obtain a grid of quadrilaterals and hexagons with the same number of matchings . ) we get @xmath313 this gives us the infinite graph of figure [ blum ] . a quick induction then shows that @xmath314 as described in the introduction , a combinatorial formula for the octahedron recurrence will give rise to a combinatorial interpretation of the somos sequences . specifically , if we set @xmath315 then the number of terms of @xmath316 will obey the somos-4 recurrence . similarly , if @xmath317 then the number of terms of @xmath316 will obey the somos-5 recurrence . figures [ somos4 ] and [ somos4finite ] show the infinite graph and the first several finite graphs corresponding to the somos-4 recurrence . figures [ somos5 ] and [ somos5finite ] do the same for somos-5 . it may be hard to see the periodicity of the @xmath178 for @xmath318 ; the tiles repeat as one travels five cells to the right , or one cell up and two to the right . in this section , we will prove the correctness of the crosses - and - wrenches algorithm . our basic strategy is to hold the point @xmath117 where we are evaluating @xmath34 fixed , while varying @xmath127 . our proof is by induction on the number of points in @xmath319 . in this section , we introduce an alternative way of viewing @xmath23 as counting the number of matchings of an infinite graph , subject to a `` condition at infinity '' to be described later . this means that _ a priori _ @xmath23 could contain an infinite number of terms , although in fact it will not , and describing the terms of @xmath23 requires an _ a priori _ infinite amount of information . on the other hand , this new method will no longer require the use of open faces and will remove the need to treat faces in the boundary as a special case in our proof . we will use this interpretation in our first proof of the main theorem . let @xmath127 be a height function . fix a triple @xmath320 , we will be interested in @xmath26 . let @xmath27 be the @xmath260 . set @xmath321 it is easy to check that @xmath322 is a pseudo - height function . we can use the method of crosses and wrenches to associate an infinite graph @xmath323 to @xmath322 . note that if @xmath12 is a face of @xmath27 then @xmath324 . thus , @xmath27 is a subgraph of @xmath323 . upon removing @xmath27 , what remains looks like figure [ infcomp ] . let @xmath325 . ] there is a unique matching ( indicated in thick lines ) of @xmath326 given by taking the middle edge of every wrench . call this matching @xmath327 . clearly , matchings of @xmath323 that coincide with @xmath327 on @xmath326 are in bijection with matchings of @xmath27 . moreover , one can easily check that the face exponents associated to a given matching of @xmath27 are the same as for the corresponding matching of @xmath323 , including that the exponent of any face of @xmath323 that does not correspond to a face of @xmath27 is 0 . thus , we could define @xmath26 as the sum over matchings @xmath28 of @xmath323 such that @xmath28 coincides with @xmath327 on @xmath326 of @xmath56 . we claim that the following , less obvious result also holds : let @xmath28 be a matching of @xmath323 in which all but finitely many vertices are matched to the same vertex as in @xmath327 . ( @xmath327 is a matching of @xmath326 , so for all but finitely many vertices of @xmath323 this makes sense . ) then @xmath28 coincides with @xmath327 everywhere on @xmath326 . suppose the opposite . all the vertices of @xmath328 come from wrenches . each wrench has one of its vertices closer to @xmath42 than the other ; call these the _ near _ and _ far _ vertex of the wrench . now , suppose the theorem is false . of the finite number of vertices of @xmath329 not matched as in @xmath28 , let @xmath274 be the furthest from @xmath42 . we derive a contradiction in the two possible cases . case 1 : @xmath274 is a near vertex . then the far vertex @xmath275 in the same wrench as @xmath274 is matched as in @xmath28 . but @xmath275 is matched with @xmath274 in @xmath28 , so @xmath275 is matched with @xmath274 and @xmath274 is matched as in @xmath327 . case 2 : @xmath275 is a far vertex . but all of @xmath274 s neighbors except the one it is matched to in @xmath327 are farther from @xmath42 than @xmath274 is , so they are matched as in @xmath327 and can not be matched to @xmath274 . so @xmath274 is matched as in @xmath327 . as result , we can give another statement of the main theorem . call a matching @xmath28 of @xmath323 an acceptable matching if @xmath28 coincides with @xmath327 for all but finitely many edges . then @xmath330 [ split ] consider a @xmath27 . replace a vertex @xmath27 with two unweighted edges as in figure [ vertsplit ] to form a new @xmath159 . then @xmath331 . any matching of the old graph can be uniquely transformed to a matching of the new graph by adding one of the two new edges . the additional edge in the matching contributes a factor of 1 in the product of the edge variables . the two faces adjacent to the new edges have one more used edge and one more unused edge than the corresponding faces of the previous matching , so they have the same exponent . the other faces are unaffected . suppose a @xmath27 contains the sub - shown in the left of figure [ ur ] with the indicated edge weights and face weights . ( the face weights are in uppercase , the edge weights in lower case . ) create a new graph @xmath159 by replacing this with the in the right of figure [ ur ] where @xmath332 then @xmath333 . this theorem is related to lemma 2.5 of @xcite . however , because he does nt use face weights , ciucu s theorem involves slightly different replacements and yields a relation of the form @xmath334 . write @xmath335 where @xmath336 consists of the terms corresponding to matchings where none of the edges @xmath83 , @xmath84 , @xmath95 or @xmath96 is used , @xmath337 consists of the terms using one such edge and @xmath338 consists of the terms using two . write @xmath339 similarly . we will show that @xmath340 , @xmath341 and @xmath342 . we first give a bijection between the matchings counted by @xmath337 and those counted by @xmath343 as shown in figure [ biject1 ] . case ] in each case , it is easy to check that the two matchings paired off have exponent @xmath9 on @xmath344 and @xmath345 and raise all other variables to the same exponent . next , as shown in figure [ biject0 ] , we give a bijection that associates to each matching counted by @xmath336 two of the type counted by @xmath346 . case ] the first transformation in figure [ biject0 ] increases the exponents of @xmath347 and @xmath348 by 1 , adds the edges @xmath83 and @xmath96 and changes the contribution of the center face from @xmath344 to @xmath349 . similarly , the second transformation increases the exponents of @xmath350 and @xmath351 by 1 , adds the edges @xmath84 and @xmath95 and changes the contribution of the center face from @xmath344 to @xmath349 . so we get from @xmath336 to @xmath346 by multiplying by @xmath352 but , as @xmath353 , this is 1 . finally , we show @xmath342 . this is similar to the preceding paragraph . this time , we pair off every two matchings in @xmath338 with one matching in @xmath354 , as shown in figure [ biject2 ] . case ] in the first pairing of figure [ biject2 ] , we delete edges @xmath83 and @xmath96 , reduce the exponents of @xmath347 and @xmath348 by 1 and replace @xmath355 by @xmath345 . in the second , we delete edges @xmath84 and @xmath95 , reduce the exponents of @xmath350 and @xmath351 by 1 and replace @xmath355 by @xmath345 . thus , we are multiplying by @xmath356 which is 1 as before . let @xmath357 and let @xmath127 and @xmath322 be as in subsection [ infext ] ; let @xmath323 be the associated to @xmath322 . our proof is by induction on @xmath358 . if @xmath359 then @xmath360 so @xmath323 is the graph shown in figure [ basecase ] which has only the indicated admissible matching . the matching polynomial of this matching is @xmath361 . we also have in this case that @xmath362 so we have @xmath363 . if @xmath364 then , among the @xmath38 for which @xmath272 there is ( at least ) one with @xmath365 minimal , let this be @xmath38 . we claim that @xmath366 . we deal with the case of @xmath367 , the other three cases are similar . we have @xmath368 as @xmath322 is a pseudo - height function . if @xmath369 , then @xmath370 contradicting the minimality of @xmath38 . thus , we have shown @xmath366 . so the face associated to @xmath205 is a square . apply the urban renewal theorem to this square to create a new graph with the same matching polynomial . then apply lemma [ split ] to this graph wherever possible . the graph thus created will be the same as replacing any crosses whose vertex lies on the face @xmath38 by a wrench with one vertex on @xmath38 and _ vice versa_. also , the edge weights adjacent to @xmath38 are interchanged with their diametric opposites and the weight @xmath24 is replaced by a certain binomial . the graph thus produced is precisely the graph produced by the function @xmath371 where @xmath372 and @xmath373 everywhere else . by induction , the matching polynomial of this graph is @xmath26 for the initial conditions @xmath371 . the @xmath34 for the initial conditions @xmath127 is given by replacing @xmath24 by the same binomial as before . thus , @xmath26 is precisely given by the matching polynomial of the graph for @xmath371 with @xmath24 replaced by the said binomial , which by the urban renewal theorem is precisely the matching polynomial of the graph for @xmath127 . in this section , we give a second proof of the main theorem , similar to the proof of @xcite . in order to avoid lengthy tedious verifications , we will perform this proof with the face variables set equal to 1 throughout this section . we first give some graph theoretic notations . let @xmath27 be a graph and @xmath374 . let @xmath375 denote the elements of @xmath277 that border vertices of @xmath27 not in @xmath277 . if @xmath27 is bipartite , colored black and white , let @xmath376 denote the number of black vertices of @xmath277 minus the number of white vertices . while @xmath376 is only defined up to sign , we adopt the implicit assumption that , if @xmath377 and @xmath378 are both subsets of @xmath154 , then @xmath379 and @xmath380 are computed from the same coloring of @xmath27 . let @xmath381 be the subgraph of @xmath27 induced by @xmath277 . we abuse notation by writing @xmath382 to mean @xmath383 . ( recall that all face variables have been set equal to 1 , so @xmath382 is a polynomial in the edge variables of @xmath381 . ) the following theorem is essentially due to kuo and proven in @xcite . ( kuo s paper proves this relation for many specific families of graphs but never states the general theorem . the below is my attempt to assimilate all of kuo s cases into a general framework which will also encompass the results of this paper . ) let @xmath27 be a bipartite planar graph and let the vertices of @xmath27 be partitioned into nine sets @xmath384 assume that only vertices in the sets joined in figure [ condense ] border each other . ( vertices may also border other vertices in the same set . ) assume further that @xmath385 and @xmath386 finally , assume that @xmath387 and @xmath388 are entirely black and @xmath389 and @xmath390 are entirely white . then we have @xmath391 we write @xmath344 to denote the collection of sets @xmath392 . let @xmath27 always denote a graph obeying the above hypotheses . we define a _ northern join _ to be a set @xmath393 of edges of @xmath27 of any of the following types 1 . a path with one endpoint in @xmath394 , the other endpoint in @xmath395 and the intermediate vertices in @xmath396 . 2 . a single edge with one endpoint in @xmath394 and the other in @xmath395 . 3 . a pair of edges , one joining @xmath394 to @xmath397 and the other joining @xmath395 to @xmath397 . we define an eastern , southern or western join analogously . let @xmath28 be a multi - set of edges of @xmath27 such that every member of @xmath396 lies on two edges of @xmath28 and every other vertex lies on a single edge of @xmath28 . then we can write @xmath28 uniquely as a vertex disjoint union @xmath398 where 1 . either @xmath399 is a northern join and @xmath400 is a southern join , or @xmath399 is a eastern join and @xmath400 is a western join . 2 . @xmath401 is a disjoint union of cycles , entirely contained in @xmath396 . ( we count a doubled edge as a cycle of length 2 . ) @xmath402 is a vertex disjoint set of edges contained entirely within @xmath403 . since every vertex is adjacent to either one or two elements of @xmath28 , @xmath28 can be written uniquely as a disjoint union of cycles and paths . moreover , the vertices of @xmath396 are on cycles or are inner vertices of paths and the vertices of @xmath404 are the ends of paths . let @xmath401 be the cycles of @xmath28 , @xmath401 is vertex disjoint from the other edges of @xmath28 . after deleting @xmath401 , the remaining graph is a disjoint union of paths , each of whose endpoints lie in one of the @xmath405 and all of whose interior vertices must lie in @xmath396 . let @xmath402 be the isolated edges connecting one point of @xmath403 to another . after deleting all of the @xmath402 , what remains are paths as before which either have at least one interior point or whose endpoints lie in different @xmath405 . call the set of remaining edges @xmath406 . let @xmath405 . we claim @xmath407 . this is because , if @xmath408 , then there is a single edge @xmath409 of @xmath28 containing @xmath274 . we have @xmath410 by the definition of @xmath411 . then @xmath409 is also the only edge containing @xmath275 and thus @xmath409 is an isolated edge of @xmath28 lying entirely in @xmath403 . so @xmath412 and @xmath413 . we know @xmath414 contains equally many black as white vertices as it is a disjoint union of edges . thus @xmath415 has one more black vertex than white , as we assumed @xmath416 . however , we just showed @xmath417 , and @xmath387 is entirely black . so there is a unique vertex @xmath418 in @xmath415 . similarly , there are unique vertices @xmath419 , @xmath420 and @xmath421 , with @xmath418 and @xmath421 black and @xmath419 and @xmath420 white . also by the same logic , @xmath422 is even for @xmath423 . now , the @xmath424 must be endpoints of paths . the _ a priori _ possible paths are the following , and their rotations and reflections : 1 . a path from @xmath418 to @xmath419 through @xmath396 . ( there are four possibilities in this symmetry class . 2 . a path from @xmath418 to @xmath421 through @xmath396 . ( there are two possibilities in this symmetry class . ) 3 . a single edge from @xmath418 to @xmath419 . ( there are four possibilities in this symmetry class . ) 4 . a single edge from @xmath418 to @xmath425 . ( there are eight possibilities in this symmetry class . ) as @xmath426 is even , there must either be 0 or 2 paths of type ( 4 ) ending in @xmath397 , and similarly for @xmath427 , @xmath428 and @xmath429 . this means we can not have a path joining @xmath418 to @xmath421 . if such a path existed , @xmath419 could not be joined to @xmath425 or @xmath347 , as that would create one path ending in that set . also , @xmath419 could not be joined to @xmath420 , as the paths @xmath430 and @xmath431 would cross . similarly , we may not have a path joining @xmath419 to @xmath420 . so the paths of type 2 do not occur . it is now easy to see that @xmath406 must decompose either as the union of a northern and a southern join or as the union of an eastern and a western join . note that , if @xmath432 ( @xmath433 or @xmath44 ) passes through @xmath396 , it will contain an odd number of edges , as its endpoints are of opposite colors . we now prove the result . each side of the equation is a sum of products of edge variables , and each product corresponds to an @xmath28 meeting the conditions of the lemma ; we will abuse notation and call this product @xmath28 as well . we will show each @xmath28 occurs with the same coefficient on both sides . specifically , let @xmath82 denote the number of cycles in @xmath401 of length greater than 2 . we claim that @xmath28 appears on each side of the equation with coefficient @xmath434 . even more specifically , we claim that @xmath435 contains @xmath28 with coefficient @xmath434 . we also claim that , if @xmath399 is a northern join and @xmath400 a southern join , then @xmath436 contains @xmath28 with a coefficient of @xmath434 and @xmath437 does not contain @xmath28 . if @xmath399 is an eastern and @xmath400 a western join then we claim the reverse holds . consider first the task of determining how many times @xmath28 appears in @xmath435 . this is the same as the number of ways to decompose @xmath28 as @xmath438 with @xmath439 and @xmath440 matchings of @xmath27 and @xmath396 respectively . clearly , the edges with endpoints outside @xmath396 must lie in @xmath439 . this means all the edges of @xmath402 , @xmath405 , the edges of @xmath432 if @xmath432 does not pass through @xmath396 and the final edges of @xmath432 if @xmath432 does pass through @xmath396 . this forces the allocation of all the edges of @xmath432 because , when two edges of @xmath28 share a vertex , one edge must go in @xmath439 and one in @xmath440 . this forcing is consistent because @xmath432 has an odd number of edges and the end edges must both lie in @xmath439 . finally , we must allocate the edges of @xmath401 . all the cycles of @xmath401 are of even length as @xmath27 is bipartite . in a cycle of length greater than 2 , we may arbitrarily choose which half of its edges came from @xmath439 and which from @xmath441 . thus , we make @xmath434 choices . next , consider the coefficient of @xmath28 in @xmath436 . we must similarly write @xmath442 we claim that if @xmath399 is an eastern join and @xmath400 a western , this coefficient is 0 . there are three cases . @xmath399 is a path joining @xmath394 to @xmath443 through @xmath396 . then every edge of @xmath399 must lie in @xmath444 or in @xmath445 and which of these two it lies in must alternate . there are an odd number of edges in @xmath399 , so the final edges must lie in the same one of these two sets . but any edge with an endpoint in @xmath394 must lie in @xmath444 and any edge with an endpoint in @xmath443 must lie in @xmath446 . @xmath399 is a single edge with endpoints in @xmath394 and @xmath443 . but none of the @xmath447 s can contain such an edge . @xmath399 is a union of two disjoint edges , one connecting @xmath394 to @xmath427 and one connecting @xmath443 to @xmath427 . but neither of these edge types can lie in any of the four @xmath447 s . finally , we show that if @xmath399 is a northern join and @xmath400 a southern join that @xmath28 has coefficient @xmath434 in @xmath436 . as before , the edges of @xmath402 , @xmath405 , are uniquely allocated to one of @xmath444 , @xmath445 , @xmath448 and @xmath449 . if @xmath432 does not pass through @xmath396 , its edges are also immediately forced . if @xmath432 does pass through @xmath396 , its final edges are forced , thus forcing the others and again we have consistency as the two final edges are in the same set and @xmath432 has an odd number of edges . finally , each cycle of @xmath401 of length greater than 2 can be allocated in two ways . the case of @xmath437 is exactly analogous . let @xmath27 be a graph arising from the method of crosses and wrenches for some height function @xmath127 and some @xmath450 . the purpose of this section is to describe a decomposition of the vertices of @xmath27 into nine disjoint sets . in the next section , we will show these sets obey the hypotheses of the previous section . for simplicity , we assume that @xmath451 . note that we have @xmath452 and @xmath453 ( recall that @xmath154 denotes the vertices of @xmath27 . ) thus , the four graphs @xmath454 and @xmath455 intersect as shown in the venn diagram in figure [ partition ] , where the unlabelled cells indicate empty intersections . we decompose @xmath456 into the nine sets @xmath396 , @xmath397 , @xmath394 , @xmath427 , @xmath443 , @xmath428 , @xmath457 , @xmath429 and @xmath395 as shown in the figure we now verify that the sets of vertices @xmath396 , @xmath397 , @xmath427 , @xmath429 , @xmath428 , @xmath394 , @xmath395 , @xmath443 and @xmath457 obey the hypotheses of the condensation theorem . we also show that @xmath458 , @xmath459 , @xmath460 , @xmath461 . these clearly establish the main theorem . the following lemma is clear . [ ineq ] abbreviate @xmath462 by @xmath463 . then @xmath464 ( we are using our simplifying assumption that @xmath465 to avoid a messy statement for this face . ) [ geomn ] we have @xmath458 , @xmath459 , @xmath460 and @xmath461 and @xmath466 . also , all of the vertices of @xmath427 that border vertices of @xmath394 are the same color . calling this color white , the vertices of @xmath397 bordering @xmath394 , those of @xmath429 bordering @xmath457 and those of @xmath428 bordering @xmath457 are white . the vertices of @xmath427 bordering @xmath443 , those of @xmath428 bordering @xmath443 , those of @xmath397 bordering @xmath457 and those of @xmath429 bordering @xmath457 are black . we do the case of @xmath427 , the others are similar . let @xmath467 and let @xmath38 be a closed face of @xmath468 bordering @xmath274 ( by the definition of @xmath427 , such a face exists ) ; by lemma [ ineq ] , @xmath469 . now , @xmath38 is not a closed face of @xmath470 . thus , by lemma [ ineq ] , @xmath471 . but , similarly , @xmath279 so @xmath472 . moreover , the faces that @xmath274 borders which are not of the form @xmath473 must not lie in @xmath27 at all , so we have @xmath474 . putting this data into the crosses and wrenches algorithm , we have that @xmath427 looks like figure [ elayout ] , where the number of hexagons and the length of the dangling path can vary . the highlighted vertices are those that border @xmath394 and @xmath443 . ] clearly , @xmath427 has a unique matching @xmath28 , shown in bold . let @xmath57 be the largest value for which @xmath473 is a closed face of @xmath27 , the only weighted edge in @xmath28 is the edge separating @xmath473 and @xmath475 . as @xmath476 and @xmath477 ( as otherwise @xmath478 we be a closed face ) we see that this is the edge @xmath479 . as @xmath427 has a matching , @xmath480 . all that is left to show is that the vertices of @xmath397 that border @xmath394 and the vertices of @xmath427 that do are the same color . it is equivalent to show that the vertex of @xmath481 and that of @xmath482 closer to @xmath394 are the same color . but the path joining them is the one we proved in proposition [ boundary ] had an odd number of edges . let @xmath483 . then exactly one of the following is true . 1 . @xmath274 lies in a wrench oriented ne - sw ( i.e. , this orientation : @xmath219 ) , the other end of which is also in @xmath394 . + or 2 . @xmath274 lies on the outer perimeter of @xmath260 . let @xmath38 , @xmath283 , @xmath284 , @xmath484 be the faces surrounding @xmath274 ( @xmath274 may only be adjacent to three of them ) and let @xmath485 be the set of these four faces .. by lemma [ ineq ] , either @xmath57 or @xmath486 so @xmath282 . similarly , @xmath279 . there are several cases . first , suppose @xmath469 . then @xmath487 for @xmath488 . if @xmath489 for all @xmath488 , then @xmath274 lies on a ne - sw wrench both ends of which are in @xmath394 . if some @xmath490 then @xmath491 is not a closed face of @xmath260 and one can check that in every case @xmath274 is adjacent to @xmath491 , so @xmath274 is on the outer perimeter . now , suppose that @xmath492 . then @xmath493 . the only way that @xmath274 could be in @xmath427 then is if @xmath274 were at the ne end of a ne - sw wrench so that it did nt border @xmath38 . then @xmath283 does border @xmath274 , so @xmath494 . the only way this is possible is if @xmath495 , @xmath496 and @xmath497 . then @xmath498 and @xmath274 lies on @xmath484 , so again @xmath274 is on the perimeter . finally , suppose that @xmath499 . then @xmath500 for all @xmath488 . but then none of the @xmath267 are closed faces of @xmath260 and @xmath274 is not a vertex of @xmath260 , a contradiction . @xmath394 and @xmath457 have one more black vertex than white , and _ vice versa _ for @xmath443 and @xmath395 . we discuss the case of @xmath394 , the others are similar . pair off all of the vertices of @xmath427 for which the first case of the previous proposition holds with the other end of their wrench . this removes equal numbers of black and white vertices . what remains is a path , which we showed in proposition [ geomn ] has each end black . @xmath387 and @xmath388 are entirely black and _ vice versa _ for @xmath390 and @xmath389 . moreover , only the adjacencies permitted by the hypotheses of the condensation theorem can occur . besides the endpoints of the path in the previous proposition , @xmath387 must be made up entirely of vertices from the wrenches in the previous proposition . however , in fact , only the south - west vertex of a wrench will be able to lie in @xmath387 , and these are all the same color . this deals with the coloration of the boundaries . this immediately means that @xmath394 can not border @xmath457 . we saw in lemma [ geomn ] that @xmath394 only borders @xmath397 and @xmath427 . these and the symmetrically equivalent restrictions gives the required restrictions on what may border what . @xmath396 has equally many black and white vertices . @xmath501 . by induction on @xmath29 , we can assume that we already know that @xmath502 . in particular , @xmath396 has at least one perfect matching . we have now checked all of the conditions to apply the kuo condensation theorem . the condensation proof of the main theorem , although it involves more special cases , seems to have at least one consequence which does not follow from the urban renewal proof . in the relation @xmath503 each monomial appearing on the right hand side appears in exactly one of the two terms on the right , and appears with the coefficient @xmath434 for some @xmath82 . this just says that every combination of edge @xmath28 as in the proof of the kuo condensation theorem occurs either only in @xmath504 or only in @xmath505 and appears with coefficient @xmath434 for some @xmath82 ; this follow from the proof of the kuo condensation theorem . we are being asked to show that we can recover a matching of @xmath182 ( henceforth , @xmath27 ) from knowing which weighted edges were use in the matching . since the unweighted edges of @xmath27 are vertex disjoint , knowing which weighted edges are used determines all the edges used , and hence the matching . we are now being asked to show that we can recover a matching @xmath28 of @xmath27 from just knowing the face exponents . by the previous result , it is enough to show that we can determine whether or not a weighted edge appears in @xmath28 . recall the notation of section [ gwofs ] : for @xmath34 a face of @xmath178 , @xmath506 is the exponent of @xmath162 and , for @xmath152 an edge of @xmath178 , @xmath507 is the exponent of @xmath152 . we give a formula for @xmath508 in terms of certain @xmath506 s in the case where @xmath152 is of the form @xmath509 , similar formulas for @xmath510 , @xmath511 and @xmath512 can be found and will be stated later in the section . my thanks to gabriel carrol for suggesting this simple formulation and proof of the lemma ( in a somewhat different context . ) now , let us consider the coefficient of @xmath515 in any term of @xmath528 . since @xmath513 appears in @xmath182 , we must have @xmath529 and @xmath530 . thus , @xmath531 and no @xmath515 appears . on the other hand , @xmath516 can also be obtained from @xmath34 by substituting @xmath532 for @xmath24 for every @xmath136 such that @xmath533 and @xmath534 and substituting @xmath535 for @xmath536 . so we see that the exponent of @xmath515 is @xmath537 1 . @xmath510 , if we define @xmath541 , @xmath542 , @xmath543 and @xmath277 by comparing @xmath544 to @xmath545 . @xmath511 , if we define @xmath541 , @xmath542 , @xmath543 and @xmath277 by comparing @xmath546 to @xmath547 . @xmath512 , if we define @xmath541 , @xmath542 , @xmath543 and @xmath277 by comparing @xmath548 to @xmath549 . a geometrical description can be given for these rules . let @xmath152 be a weighted edge of @xmath27 . draw four paths in @xmath27 by starting at @xmath152 and alternately turning left and right . ( there are two ways to leave @xmath152 and two choices for which turn to make first . ) these divide the faces of @xmath27 into four regions @xmath541 , @xmath542 , @xmath543 and @xmath277 . if the edge @xmath152 is present , then the sums @xmath550 , @xmath551 , @xmath552 and @xmath553 will yield a 1 , two -1 s and a 0 . if @xmath152 is absent , one of these sums will be 1 and the other three will be 0 . interestingly , if one follows the operation of the previous where @xmath152 is an unweighted edge , one gets precisely the same conclusions except that the interpretation of the results is changes : getting the sums @xmath554 ( in some order ) now means that @xmath152 is _ present _ and @xmath555 means @xmath152 is _ let @xmath27 be a bipartite connected planar graph with a fixed labeling of its vertices as black and white and let @xmath275 be a function assigning a real number to each edge of @xmath27 . we define a _ propp height _ to be a function @xmath556 from the faces of @xmath27 to the real numbers with the following property : let @xmath34 and @xmath557 be faces of @xmath27 separated by an edge @xmath152 such that , when one stands on @xmath34 and looks toward @xmath557 , the white vertex of @xmath152 is on the left . then @xmath558 . define two propp heights @xmath556 and @xmath559 to be equivalent if @xmath560 for some constant @xmath95 independent of @xmath34 . ( the term propp height is my own , to distinguish them from the heights which occur through out this paper . ) for @xmath27 as above , one can find @xmath275 such that there is a bijection between equivalence classes of propp heights for @xmath275 and matchings of @xmath27 . the bijection is such that , if @xmath556 corresponds to @xmath28 , the edge @xmath152 occurs in @xmath28 iff @xmath561 where @xmath152 separates @xmath34 and @xmath557 . these propp heights have been very useful in studying statistical properties of random matchings , see _ e.g. _ @xcite . for crosses and wrenches graphs , the propp height has an extremely simple description . let @xmath178 be a crosses and wrenches graph . we may take @xmath275 in the previous theorem to be given by @xmath562 , if @xmath152 is a horizontal or vertical edge and @xmath563 , if @xmath152 is a diagonal edge . let @xmath27 be a crosses and wrenches graph . suppose that we want to randomly generate a matching of @xmath27 , with our sample drawn from all matchings of @xmath27 with uniform probability . such sampling has produced intriguing results and conjectures in the cases of aztec diamonds and fortresses . it is possible to do so in time @xmath566 , where @xmath567 can be any of @xmath568 , @xmath569 or @xmath570 as all of these only differ by a constant factor for @xmath27 a crosses and wrenches graph . more specifically , if @xmath571 , one can do so in time @xmath572 . ( in most practical cases , this is closer to @xmath573 . we give only a quick sketch ; the method is a simple adaptation of the methods of @xcite . let @xmath571 arise from a height function @xmath127 and let @xmath574 . throughout our algorithm , @xmath131 will denote a function @xmath575 . our algorithm takes as input @xmath127 and @xmath131 and returns a random matching , where the probability of a matching @xmath28 being returned is proportional to @xmath576 . we describe our output as a list of the weighted edges used in @xmath28 . at the beginning of the algorithm , we take @xmath577 for all @xmath34 . * step 5 : * for shorthand , set @xmath589 , @xmath590 , @xmath591 and @xmath592 . if @xmath588 contains one of @xmath593 , return m and halt . if @xmath588 contains two of @xmath593 , return @xmath594 and halt . if @xmath588 contains none of @xmath593 , return @xmath595 with probability @xmath596 and return @xmath597 with probability @xmath598 . all other cases are impossible . in order to not do a time consuming search at step 2 , one can keep a reverse look - up table which , given @xmath127 , record all @xmath38 with @xmath599 and is updated in step 3 when @xmath127 is replaced by @xmath322 . note that @xmath600 will always be integer valued	we study a recurrence defined on a three dimensional lattice and prove that its values are laurent polynomials in the initial conditions with all coefficients equal to one . this recurrence was studied by propp and by fomin and zelivinsky . fomin and zelivinsky were able to prove laurentness and conjectured that the coefficients were 1 . our proof establishes a bijection between the terms of the laurent polynomial and the perfect matchings of certain graphs , generalizing the theory of aztec diamonds . in particular , this shows that the coefficients of this polynomial , and polynomials obtained by specializing its variables , are positive , a conjecture of fomin and zelevinsky .
in a surprising result , hartle and hawking twenty years ago showed @xcite how , at least in the semi - classical theory of quantum gravity , one could give an intuitively appealing definition for a preferred wave function for the universe . if one believes that nature selects this definition , then they argued that this wave function explains various features which we observe the universe to have . one of these features is that the universe is simple when it is small @xcite and becomes more and more complex as it grows in size . while hawking originally believed that the universe must always be simple when small @xcite , page and laflamme @xcite pointed out that this wave - function would be expected to also contain probablilities for the universe to be complex when small . this was interpreted as indicating a classical history of the universe which began simple when small , expanded , with the gravitational instability causing growth in the quantum fluctuations present when small . the universe would eventually re - collapse with the fluctuations continuing to grow . if one wrote the hh wave - function in terms of a basis which contained such semi - classical components , one would interpret them to say that the thermodynamic arrow of time would be driven by these fluctuations and that one should interpret those solutions such that the low fluctuation end should be interpreted as the beginning and the high as the end of the evolution . however , there is an alternative interpretation of the hh wave - function ( see for example jheeta and unruh@xcite ) . the prescription bears a strong resemblance to the prescription of a semi - classical wkb wave - function near a classical turning point of a potential . the classical solution for the hh models is that of a de sitter space with the the cosmological constant dominating the evolution ie the classical solution looks like de sitter space in spherical coordinates near the point where the spherical space reaches it minimum size . interpreted in this way there is another place where the universe could and does bounce ; namely if a closed universe has matter such that the energy density of the matter drops as the universe expands . in this case , the curvature terms in the frw equations eventually dominates and halts the expansion and the universe re - collapses . the question which we address is whether the hh prescription could be used to place boundary conditions at this bounce point rather than at the point where the universe achieves it minimum size ? we will examine a simple model , with a homogeneous scalar field driving the dynamics of the universe . the space - time will be assumed to be a closed spherical homogeneous and isotropic frw cosmology with metric ds^2= n^2 dt^2 - a(t)^2 ( dr^2 + sin(r)^2 ( d^2 + sin()^2 d^2 ) ) the mini - superspace action for this space - time becomes s= n a^3 ( -6a^2n^2 a^2 + 6a^2 + 12n^2 ^ 2 -v ( ) ) dt which has as semi - classical equations of motion + na^3dv()d=0 + 12adndt dandt+6 dandt^2 + 6 + 3a^2(12n^2 ^2 - v)=0 + ( a ^22 n^2 a^2 + 6a^2 - 12n^2 ^ 2 -v ( ) ) = 0 where @xmath0 . ( the third is the constraint equation , and is an integral of motion of the other two ) . the semi - classical hh prescription is that the path integral will be dominated by a solution of the equations which has only a final boundary given by the classical state on which one wishes to evaluate the wave function , ( i.e. , the single boundary is defined by the values of @xmath1 and @xmath2 ( or @xmath3 ) which one wants on the final boundary ) . we must find solutions to these equations which have no initial boundary which have @xmath4 going to zero and @xmath5 regular as @xmath4 approaches zero . this is impossible if @xmath6 ( and @xmath1 ) are real . we can however find such solutions if @xmath6 is chosen to be complex , and in particular if we choose @xmath6 to become purely imaginary as @xmath1 approaches zero . ( for a generalisation to arbitrary complex @xmath6 see jheeta and unruh @xcite ) . one choice , if it works , is to take @xmath6 and @xmath1 to be purely real over some range of times ( the lorentzian regime ) , and then where @xmath4 has zero derivative , to take @xmath6 to be purely imaginary . this allows one to smoothly join @xmath4 across the surface where @xmath6 discontinuously . however , in general , both @xmath7 and @xmath3 can not be taken to be zero at the same time @xmath8 and still find a solution with only a single boundary . since the equations of motion for @xmath1 and @xmath2 are smooth second order equations , the derivatives @xmath9 must also be smooth . while we can keep @xmath4 real across this boundary , since going from real zero to imaginary zero is a continuous transformation , we can not do the same for @xmath2 . thus , @xmath2 must thus become complex as one crosses this boundary . we are going to try to model the universe such that at time @xmath10 we have @xmath11 , and for @xmath12 , the universe contracts from the value of @xmath1 at this time . ie , we take @xmath10 to be the time at which the universe achieves its maximum value in the lorentzian regime . examining the constraint equation for @xmath1 , let us examine this boundary where @xmath13 . since @xmath7 is , by assumption less , than zero ( the universe is contracting ) for larger values of @xmath8 , the potential term must be sub - dominant . the curvature term @xmath14 cancels with the kinetic energy . if @xmath15 were to dominate , then @xmath16 decreases as @xmath1 decreases , and @xmath17 would become imaginary . thus , in order that we have such a maximum in the radius of the universe , the term @xmath18 must be dominating the energy , and @xmath19 must be decreasing at least as fast as @xmath20 but for very small @xmath15 this is precisely how @xmath21 is doing . ( for small @xmath22 , @xmath2 obeys 0 or ie , in order that the universe re - collapse , the potential term can not dominate the evolution of the universe it must be dominated by the kinetic energy of the matter . now , @xmath23 must be continuous across that boundary where @xmath6 goes from real to imaginary . since we are taking n to be purely imaginary for @xmath24 , @xmath21 must also be imaginary for times just before zero . ie , @xmath2 must pick up an imaginary part . let us now examine the evolution of this model in this realm where @xmath6 is purely imaginary . let us assume that the potential @xmath25 is such that the potential remains real for @xmath26 , where @xmath27 is the value of @xmath2 at the point where @xmath11 . since @xmath6 is arbitrary , we will take @xmath28 for @xmath12 and @xmath29 for @xmath24 . the equation of motion for @xmath30 will be a^3 ddt = d vd and for @xmath1 , the constraint equation gives ( dadt)^2 = -1 + 16a^2(12ddt^2 + v ( ) ) the second order einstein equation for @xmath1 shows that through @xmath10 , @xmath31 reverses sign . thus , while the universe contracts toward the future in the lorentzian regime , it must expand into the past at this turning point . thus @xmath32 is less than zero for @xmath24 near 0 , and the universe expands as go to negative @xmath8 . this raises the question of how one could get @xmath1 to go to zero in this n - imaginary region . the kinetic energy which dominates is driving the universe larger . the answer is that as @xmath1 increases , @xmath33 keeps decreasing . if the potential @xmath34 remains non - zero , eventually it will dominate once again . once it dominates , the potential term will keep increasing until it becomes of the same order as the curvature term ( the 1 in the equation for @xmath7 ) and @xmath1 has another turning point . the universe will then collapse as a function of @xmath8 , the kinetic term will again grow , and the danger is that one will hit another turning point . if however the potential is arranged so that during this collapse , the potential increases , and thus @xmath35 decreases , one can arrange the system so that @xmath1 can collapse all the way to zero smoothly . let us examine a specific model . we take the potential to be v()= v_0 + ^4 . we are going to assume that at t=0 , @xmath36 , @xmath37 is large , and @xmath38 . we will assume that @xmath39 is smaller than @xmath40 and thus @xmath41 . if the solution is such that @xmath1 approaches zero smoothly at @xmath42 ( so that @xmath30 and @xmath33 are both finite there ) then near @xmath43 we have ( t)(t_0 ) ( 1 - 8(t - t_0)^2 + ... ) + a(t)=(t - t_0)(1- ( v_0 - 4(t_0)^4)(t - t_0)^2 + ... ) in figure 1 we plot a solution with @xmath44 and @xmath45 . the lower axis is note that the field @xmath30 approaches a constant for most of the evolution . if the potential in the imaginary direction is sufficiently flat for a large enough value of @xmath30 , the universe can oscillate between the the kinetic @xmath33 dominated bounce at smaller values of @xmath1 and the potential dominated one at large values of @xmath1 for many bounces . in figure 2 we have a solution with @xmath46 and @xmath47 , where we have three bounces before the radius eventually goes to zero . in the @xmath6 real regime , the potential @xmath1 will decrease , and the kinetic dominated term will grow roughly as @xmath48 . if the kinetic energy dominates , the solution for @xmath1 goes as @xmath49 , and @xmath2 grows logarithmically . if the potential in the real @xmath2 direction grows , it may eventually dominate the matter again , and the universe would suffer a de sitter like bounce at small values of @xmath1 . figure 2 is the solution for @xmath6 and @xmath2 real into the past from the large radius bounce which matches to the solution for figure 1 . in this particular case , because the potential eventually grows so rapidly with @xmath2 , the smaller bounce radius is not that much smaller than the largest size . however this depends entirely on the behaviour of the potential for large real @xmath2 . thus , if our potential were say @xmath50 , with @xmath51 very large , our potential along the real @xmath2 axis would remain bounded by @xmath27 , and the dynamics of @xmath1 would be kinetic dominated for all times to the past of the large bounce . in the imaginary direction , @xmath25 would have the required confining shape and would , once @xmath2 had grown to of order @xmath51 , allow for the radius to go smoothly to zero . in the real @xmath2 and @xmath6 regime , the solution would be kinetic dominated all the way to a singularity . in the spirit of hh , we would interpret this universe as beginning in the euclidean regime with @xmath52 and @xmath53 and @xmath3 regular . the universe would `` expand '' and bounce between the potential and kinetic dominated bounce points for as long as desired , until it suddenly makes the transition to real @xmath6 and real @xmath2 at the lower kinetic dominated bounce point . the universe would then collapse into the future , until finally , either the potential dominated and the universe suffers a de sitter , inflationary type bounce , or until the the universe finally collapses to a singularity . since these solutions are possible , how would they enter into the path integral for the wave function of the universe " ? there has of course been much heat expended on the various schools as to the sign one should take for the action in the euclidean regime . these amount to either a choice of sign for @xmath6 in the imaginary regime , or a choice in the sign of @xmath1 ( since @xmath1 enters only quadratically in all expressions of physical relevance ) . thus the contributions of these imaginary @xmath6 parts of the action will contribute either @xmath54 or @xmath55 where @xmath56 is the integrated action for the imaginary @xmath6 regime ( or is the imaginary part of the action in the case of a generic complex metric and field dominating the action ) . for our purposes , this will be the value of s_i= ( -6a + aa ^2 + a^3(12 ^ 2 + v(i ) ) ) dt during the bouncing , all the terms are of the same order as the first term , and we have s_ia dt by hypothesis both @xmath1 and @xmath8 are large , and furthermore , @xmath8 in general will be of the same order as @xmath57 , where @xmath58 is the number of bounces in the imaginary @xmath6 regime , making @xmath59 of order of @xmath60 . @xmath1 here will be of order its maximum where @xmath61 . ie , the action will be of order @xmath62 . for small @xmath39 , and for the positive choice of the @xmath56 in the action , this would completely dominate the path integral . the theory would predict with overwhelming probability that the universe began large and then collapsed down to a small size , rather than the more traditional way around . furthermore , the fluctuations would grow during the collapse phase . the universe would grow more and more chaotic as its size decreased , with the simplest being when the universe was at its largest diameter . this sign is the sign usually attributed to hartle and hawking @xcite . the alternative sign , usually attributed to linde and vilenkin @xcite , @xcite ( although derivable as simply one of two possibilities in defining the hartle hawking wave function ) would on the other hand make the contributions of this particular solution to the semi - classical equations , completely sub - dominant . one of the interesting features of this analysis is the realization that the in order to maintain continuity of the scalar field equation through the transition from the real to imaginary @xmath6 , one must also allow @xmath2 to become complex . in particular , if the field and scale factor are real in the real @xmath6 regime , then the derivative of the field in general must be pure imaginary in the imaginary @xmath6 regime . thus , we note that the sign of the kinetic term @xmath63 does not change on the transition from real to imaginary @xmath2 . let us examine the action for the inhomogeneous minimally coupled scalar field @xmath64 in the geometry . the action for this field is s= na^3 ( 12n^2 ^ 2 - 1a^2()^2)dt d^3 where the @xmath65 is the spatial integral , and @xmath66 represents the spatial derivative part of the action . again at the surface where @xmath6 makes the transition from real to imaginary , and @xmath7 is zero , @xmath64 must again be complex to keep @xmath67 continuous . ie , again we can write @xmath68 . the action then becomes , taking @xmath29 , s_e= ia^3(12 ^2 + 1a^2()^2 ) dtd^3 taking the semi - classical approximation to the path integral with the vl sign we have @xmath69 as the contribution to the wave - function . all the terms in the action are positive , indicating that we want to minimize the action to find the dominant terms . using the apherical eigenstates for the spatial modes of the field we have the wave function factor as e^ - a^3 ( 12_k^2 + k^2a^2 _ k^2 ) dt near @xmath52 , a will again be linear in @xmath8 , and the solution for the field which is regular near @xmath43 is _ 0(1 + 18 t^2 + ... the action exponent can be written as a^3 ( 12_k^2 + k^2a^2 _ k^2 ) dt = a^3_k(t)_k(t)_a=0^t the lower limit is zero . as long as @xmath70 @xmath71 will slow roll " and _ k_k while in the other limit , _ k(ka)_k thus , at the transition point , the exponent will be proportional to @xmath72 ( ie quadratic ) and will be concentrated near zero with a width proportional to either @xmath73 for short wavelengths , or @xmath74 for long wavelengths . hartle and s.w . hawking , rev d 28 , 2960 ( 1983 ) . j. halliwell and s.w . hawking , d 31 , 1777 ( 1985 ) . s.w . hawking , phys . d 32 , 2489 ( 1985 ) . d.n . page , phys . rev . d 32 , 2496 ( 1985 ) . m. jheeta , w. unruh http://arxiv.org : gr - qc/9812017 `` complex paths and the hartle hawking wave - function for slow roll cosmologies ''	using the extension of the standard hawking - hartle prescription for defining a wave function for the universe , we show that it is possible , given a suitable form for the scalar field potential , to have the universe begin at its largest size and thereafter contract , with the growth of perturbations proceeding from small , at the largest size , to largest when the universe is small . this solution would dominate the wave function by an exponentially large amount if one chooses the hartle hawking prescription for the wave - function , but is exponentially sub - dominant for the linde - vilenkin prescription . # 1#2 figure # 1 _ # 2 _ epsf
the virgo cluster represents an ideal laboratory for the study of galaxy evolution in a cluster environment . it is one of the nearest galaxy clusters ( d@xmath217 mpc ) , a relatively populous system ( abell richness class i ) , consisting of some two thousand catalogued members in the virgo cluster catalogue ( vcc , * ? ? ? the virgo cluster is spiral - rich and dynamically young , i.e. , made of galaxy subgroups which are falling into the main cluster . the most prominent groups are those associated with m86 to the west and m49 to the south . a particularity of the virgo cluster is that the distribution of its hot intracluster medium is strongly peaked on the central cd galaxy m87 @xcite , and overall highly sub - structured . the virgo cluster is therefore ideal to study the effects of both gravitational interaction and interaction with the intracluster medium on the atomic gas ( ) disks of spiral members . morphological signs of tidal interaction are usually detected in the restframe ultraviolet or blue - filter optical images of the stellar disk . a series of morphological parameters , originally developed for galaxy classification , have been used extensively in recent years to identify disturbed galaxies , and to infer the merger fractions and rates over cosmological time scales @xcite . the observational benefits are that the hubble space telescope observations of high - redshift galaxies and ground - based observations of nearby galaxies ( e.g. , sdss ) probe the same wavelength , physical structure and volumes . however , in the nearby universe , the 21 cm perspective on galaxies is set to catch up to the optical surveys in terms of numbers of galaxies studied . currently under construction are two new radio observatories : the south african karoo array telescope ( meerkat ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and the australian ska pathfinder ( askap ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? together with refurbished observatories such as the extended very large array ( evla ; * ? ? ? * ) and the aperture tile in focus instrument ( apertif ; * ? ? ? * ; * ? ? ? * ) on the westerbork synthesis radio telescope ( wsrt ) , detailed maps of thousands of galaxies will become available . the unprecedented volume of data implies that the data reduction and subsequent morphological classification can only be done through automated procedures . in the previous papers in this series @xcite , we have shown that the morphological classifiers that are in use to identify mergers in the optical can be extremely useful for the perspective . the observations are not as high - resolution as the optical disk but this is compensated for by the larger extent of the atomic gas disk and the relative sensitivity of gas to an interaction for which much anecdotal evidence already existed in the rogues galaxies " catalogue @xcite . the sensitivity of surveys to gas - rich and minor mergers will provide the calibration for higher redshifts where these types of mergers are expected to be the dominant type of galaxy interaction @xcite . the square kilometre array @xcite will subsequently be able to resolve disks at these higher redshifts . the disks of spiral galaxies are also extremely sensitive to ram pressure by the intergalactic or intracluster medium . the virgo cluster is an ideal laboratory to test the effects of both gravitational interaction and intracluster medium stripping on the quantified morphology of disks . there are detailed models of the gas stripping sequence @xcite , and a uniform data set from the vla imaging of virgo spirals in atomic gas ( viva ) project @xcite , with detailed notes on individual galaxies identifying morphological phenomena . in this paper , we explore the distribution of the morphological parameters from the viva column density maps in relation to estimates by @xcite , @xcite or @xcite of the star - formation rate , deficiency , stripping and gravitational harassment . first we discuss data and morphological parameters ( [ s : data ] and [ s : param ] ) , but readers already familiar with these can skip to results and conclusions ( [ s : results ] and [ s : concl ] ) . the viva project @xcite has compiled the largest high - resolution imaging database of the virgo cluster to date . they have published 48 vla data cubes and maps with 51 virgo cluster spiral galaxies . these are some 40 new observations with the remainder from reprocessed archival data . we obtained the zero order moment maps from http://www.astro.yale.edu/viva/ and converted to column density maps using the expressions in @xcite . the spatial resolution of viva data is approximately 15@xmath3 ( or @xmath21.2 kpc at the distance of virgo ) , and the typical sensitivity is @xmath4 @xmath5 in 3@xmath6 per 10 km / s channel . viva data show that ( i ) the galaxies located in the core of the cluster predominantly have truncated disks with some gas displaced from the galactic disk and ( ii ) some 7 out of the 50 sample galaxies ( @xmath210% ) show long , one - sided tails @xcite . these tails occur in galaxies located at intermediate projected distances between 0.5 and 1 mpc from the cluster centre . from the sample listed in table 1 of @xcite , we did not use ic 3418 as it is a non - detection in the viva survey and holmberg vii because it does not have enough continuous flux to meaningfully compute morphological parameters . two morphological parameter schemes are now in wide use : concentration - asymmetry - smoothness ( cas , * ? ? ? * ) and gini20 @xcite . in an image with @xmath7 pixels with intensities @xmath8 at pixel position @xmath9 , cas the morphological parameters are defined as : @xmath10 @xmath11 @xmath12 where @xmath13 is the radius which includes that percentage of the intensity of the object , @xmath14 is the value of the pixel in the rotated image and @xmath15 is the same pixel in the image after smoothing . gini and 20 are defined as : @xmath16 @xmath17 with @xmath18 $ ] , where ( @xmath19,@xmath20 ) is the galaxy s central position , and @xmath21 is the intensity of pixel @xmath22 in an flux - ordered list . pixel @xmath23 is the pixel marking the 20% point in this list . @xmath24 is the mean pixel value . the gini parameter is an indicator of inequality in the distribution of pixel values ; gini=1 is perfect inequality with one pixel containing all the flux from an object and gini=0 is perfect equality with each pixel containing an equal fraction of the flux . values in between indicate how concentrated the flux from this object is in a few bright areas . the 20 parameter is an indication whether or not the brightest parts of the object are close to the center or spread throughout the object . in addition to these well - established parameters , we compute the gini coefficient of the second order moment @xmath25 $ ] , which we called @xmath26 @xcite . for all these parameters , we compute the uncertainty by varying the input centre the resolution and randomizing the pixel values . the exception is the gini parameter as this does not rely on the centre as input . for the gini error , we jacknife the pixel selection . for a more detailed description of these parameters , we refer the reader to the previous papers in this series @xcite and the original papers introducing these @xcite . we compute these parameters for the area defined by the contour level of @xmath27 @xmath5 ( the @xmath28 level ) . in @xcite , we identified several parts of parameter space where tidally interacting galaxies reside for the whisp sample . to identify the interacting galaxies , we used a subsample of 154 spirals and dwarf galaxies in whisp , which have independent visual estimates of interaction from either @xcite or @xcite . the criteria from the literature and our previous paper are identified in figures [ f : tail ] and [ f : int ] with dashed and dotted lines respectively . criteria from the literature are : @xmath29 from @xcite , marked by the dashed lines in figures [ f : tail ] and [ f : int ] , panels iv , v and vi . @xmath30 from @xcite , marked by the dashed lines in panel ii of figures [ f : tail ] and [ f : int ] . @xmath31 @xcite , marked by the dashed lines in panel iv of figures [ f : tail ] and [ f : int ] . we defined three criteria based on the whisp morphologies : @xmath32 makred by the vertical dotted line in figures [ f : tail ] and [ f : int ] , panels i , iii , vi and x , @xmath33 marked by the dotted line in figures [ f : tail ] and [ f : int ] , panel v , and , @xmath34 which is the dotted lines in figures [ f : tail ] and [ f : int ] , panel ix . these criteria select a merger from its morphology for different time scales . in @xcite , we quantified the typical time scale for a merger to be selected by each of these criteria using maps of merger simulations ( originally presented in * ? ? ? * ; * ? ? ? these can be compared to merger simulations of an isolated disk , evolving passively . the comparison shows how long a merger is selected as well as the level of contamination by passive disks . at the resolution of the whisp sample , interactions were identifiable as such by their morphology for about a gyr . this is a similar time scale as the one for optical morphological identifications and better than the close pair method . our goal is to explore if the above criteria could be applied to data in a denser environment , where many other phenomena play a role in the morphology as well as to explore how many of those other phenomena leave a signature in the morphology . figures [ f : tail ] and [ f : int ] show the spread of morphological parameters for the entire viva sample with different morphological classifications from @xcite . table [ t : viva ] lists all the values . due to the combination of distance and resolution , the viva sample has the coarsest physical scales we have applied these morphological parameters to date . in @xcite and @xcite we showed that these parameters are relatively invariant with distances out to d@xmath220 mpc and a resolution of 6 " , so the viva sample is effectively at the limit of this technique with current observatories . we used the optical positions ( j2000 ) of the galaxies reported in @xcite as the centre of the galaxy ( @xmath19,@xmath20 ) to compute the parameters that depend on a central position ( all except the gini paramter ) . our motivations are that the stellar disk marks the centre of the gravitational potential of a galaxy and kinematic and stellar centres generally align well ( see * ? ? ? * ; * ? ? ? optical positions of the galaxies are a likely starting point for many of the future surveys and @xcite reported many tidal tails and displacements , which could be quantified using our morphological parameters . thus , it is entirely possible that some of morphological values ( e.g. , asymmetry ) are extreme since the centre is not the barycentre of the flux , as is common for optical morphological measurements . @xcite supply extensive notes on the morphology and the nature of the stripping of individual galaxies . we have summarized these evaluations in table [ t : chung ] . @xcite note which galaxies are currently undergoing or have recently undergone stripping , as well as the likely origin of this stripping ; ram pressure by the intracluster medium or tidal interaction with a close companion . in addition , they report a parameter for deficiency ( @xmath35 ; the galaxy s surface density compared to the value typical for spirals ) , a measure of truncation ( @xmath36 ; the ratio of the diameter over the optical , b - band one ) , and additional notes for indicators of warps in the morphology and velocity maps . finally , @xcite also list the star - formation properties of their sample , as catalogued and classified by @xcite ; the level of activity and whether it was recently cut - off ( truncated star - formation ) . we have plot the spread of our morphological parameters with each of these estimates marked to explore if there are specific parts of parameter space that single out an disk characteristic , similar to the separation of interaction and non - interacting galaxies we found in the whisp database . figures [ f : hidef ] , [ f : tail ] , and [ f : int ] , show the spread in parameters each with certain phenomena marked , so that they can be identified in the morphological parameter space . we discuss several morphological phenomena noted by chung et al . below . figure [ f : hidef ] shows gini versus concentration for deficient and truncated disks . both parameters are sensitive to how evenly the flux is distributed over the area of the disk and therefore could be sensitive to both these phenomena . to identify truncation , one could use the concentration of the emission . we have found that values around or below c=1 seem to be very indicative of truncation ; the radii containing 20 and 80 percent of the flux are very close together ( fig . [ f : hidef ] , left panel ) . given that typical values for concentration in are much higher ( typically c@xmath375 , * ? ? ? * ; * ? ? ? * ) , this may be a good tracer of truncation in large surveys . a value of concentration below unity selects half of the truncated galaxies in the viva sample very cleanly ( 12 out of 24 truncated disks with no contamination from non - truncated disks ) . the deficiency is defined by the surface density compared to the average surface density for spirals ( see equation 5 in * ? ? ? the majority of viva galaxies are deficient ( 44 galaxies with @xmath38 ) . the low values of concentration ( @xmath39 ) appear to be indicative of this deficiency . in viva , those galaxies with extreme deficiency ( @xmath40 ) are not much more concentrated than moderately deficient ones ( figure [ f : hidef ] , right panel ) , at least at this resolution . we note that a c@xmath415 criterion will select some 80 % of the deficient galaxies in the viva sample but their number is contaminated ( 17% of this selection is not deficient ) . in the upcoming large surveys this makes concentration a quick assessor on whether or not a spiral has a typical content or not . nearly all the spirals in the virgo cluster are undergoing , or have recently undergone , some type of interaction . discerning between the dominant mechanism , tidal or ram - pressure , was a major focus of the work by @xcite and @xcite . several interaction related morphological phenomena ( e.g. , warps and tails ) are noted by @xcite . table [ t : chung ] lists many galaxies with tails in the viva data and several with warps . one would expect that high values for the asymmetry would identify tails ( figure [ f : tail ] ) . however , because we used the optical centres listed by @xcite , asymmetry is already at the maximum value for most of the viva galaxies ( figure [ f : tail ] , panels iv - vii ) . we note however that the asymmetry may still be a good indicator of ( tidal ) tails , but in less extreme or possibly better resolved cases . the 20-gini selection criterion ( see eq . [ eq : lcrit2 ] ) does reasonably separate those galaxies with tails from those which do not ; 60 % of the tails identified by @xcite are selected by this criterion ( figure [ f : tail ] , panel ii ) . however , these are about half the objects selected do not have an tail . therefore , the gini-20 criterion can be used a likelyhood for an tail but it will suffer from contamination . warps were identified predominantly by their kinematic signature in @xcite and no morphological parameter or combination seems to be able to identify them in the viva data . figure [ f : int ] shows the same spread of morphological parameters as figure [ f : tail ] with the estimate we gleaned from @xcite whether or not the galaxy is stripped , undergoing ram - pressure , or undergoing tidal interaction . the sequence of stripping in the virgo cluster according to @xcite is indicated by the solid lines . this sequence of six virgo galaxies does trace a track into and out of certain interaction criteria defined previously in the literature and the third paper in this series . the extreme values for asymmetry drive almost all of the viva galaxies into the asymmetry - based merger selection space ; the asymmetry criterion from @xcite , the asymmetry - gini criterion from @xcite and our asymmetry-20 criterion from @xcite ( equations [ eq : lcrit1 ] , [ eq : lcrit3 ] , and [ eq : crit2 ] respectively ) . the extreme values for concentration linked to deficiency and truncation similarly make the concentration20 selection criterion ( equation [ eq : crit3 ] , * ? ? ? * ) not applicable for merger selection . only the gini-20 criterion ( eq . [ eq : lcrit2 ] ) appears to separate objects with tails moderately as discussed above . the criterion ( eq . [ eq : crit1 ] ) is not contaminated by galaxies undergoing ram - pressure stripping but does not select tidal interactions well anymore either , something that we attribute to the native viva resolution ( no visibility times found in * ? ? ? * for the viva resolution ) . because stripping , deficiency and truncation are interlinked , there is the possibility that the concentration selection criteria also will select those galaxies that are or have been stripped . two effects move morphologies into the selection criteria : poorer resolution and additional effects such as the ram - pressure stripping . given how the vollmer sequence tracks into ( and out again ) of the merger selection criteria , we can conclude that galaxies undergoing ram - pressure stripping will sometimes be selected as interacting in a morphological selection in future surveys , at least those with spatial resolutions similar to viva . the addition of some galaxies that are undergoing stripping would , in part , explain the high merger fraction and rate we found in the whisp survey @xcite . in @xcite , we noted that the lower resolution data of the viva sample would have some problems selecting interacting systems from isolated ones , purely on the basis of their morphology ( table 8 in the appendix ) . however , the success with which tidal tails could be separated is encouraging . therefore , in future medium - resolution surveys ( e.g. , wallaby , koribalski et al , _ in preparation _ ) , the selection of disks with disturbed morphology can be used as a first pass in a search for stripping or interacting galaxies and a merger fraction and rate will have to be corrected for the contamination by stripping galaxies . for example , those galaxies that are deficient can be removed before a merger rate is determined but that would bias against denser cluster environments ( which do include more other interactions like ram - pressure stripping ) . a higher resolution survey of the virgo cluster galaxies would resolve the population of tidally interacting galaxies . the separation of ram - pressure stripped galaxies from tidal interacting ones may then be possible without the removal of deficient ones . we explored the link between the morphological parameter space of and the star - formation and truncation of star - formation as listed in table 3 in @xcite . relations are weak as the gas is undergoing several other effects other than feedback from star - formation but it is interesting to note that those galaxies with little star - formation also have lower values of gini for ( i.e. , they are more homogeneous ) . in this paper , we present the morphological parameters commonly used in the optical , applied to the column density maps of virgo cluster galaxies obtained by the viva program . we find that the resolution of these observations ( @xmath215 " or 1 kpc ) is on the coarse side for this type of analysis but some morphological phenomena are identifiable , though with limitations . a critical choice is whether or not to use the optical or the centres for these galaxies , one that every survey will have to make . the choice of centre affects measures such as concentration , 20 , , and especially asymmetry . alternatively , one could use the barycentre of the emission itself but the optical centre is most likely the centre of the dark matter halo of the galaxy and this approach would likely miss displacements like tidal tails . we find that of the morphological phenomena , deficiency and truncation can be identified easily by a low value of the concentration index ( figure [ f : hidef ] ) . warps could not be identified using the morphology parameters and tails can be identified moderately well with the combined gini and 20 interaction criterion ( eq . [ eq : lcrit2 ] & figure [ f : tail ] , panel ii ) from @xcite . galaxies that are interacting are still identified by some of the morphological criteria from the literature @xcite and those identified by us specifically for in @xcite ( eq . [ eq : lcrit1 ] [ eq : crit3 ] ) . however , galaxies affected by ram - pressure stripping are ofttimes also selected by these criteria . for example , we note that the ram pressure sequence identified by @xcite tracks into the interaction part of the parameter space for many of the morphological selection criteria ( solid tracks in figure [ f : int ] ) . discerning between the origin of the stripping ram pressure or tidal is , however , not possible with these morphological parameters , at least at viva s spatial resolution . thus any selection of interacting systems in a survey based purely on the morphology parameters will include those that are undergoing ram - pressure stripping in a denser environment . a straightforward correction would be to exclude all deficient ( @xmath39 ) galaxies from a morphological selection of interacting galaxies . we wish to thank the anonymous referee for his or her comments , helping us to greatly improve the final manuscript . we thank the viva collaboration for making their data public and easily accessible . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . we acknowledge support from the national research foundation of south africa . the work of b.w . holwerda and w.j.g . de blok is based upon research supported by the south african research chairs initiative of the department of science and technology and the national research foundation . w. j. g. , jonas j. , fanaroff b. , holwerda b. w. , bouchard a. , blyth s. , van der heyden k. , prizkal n. , 2009 , in conference proceedings of the `` panoramic radio astronomy : wide - field 1 - 2 ghz research on galaxy evolution '' , june 02 - 05 , 2009 groningen , the netherlands j. e. , van gorkom j. h. , rupen m. p. , schiminovich d. , 2001 , in astronomical society of the pacific conference series , vol . 240 , gas and galaxy evolution , hibbard j. e. , rupen m. , van gorkom j. h. , eds . , pp . 657+ b. w. , de blok w. j. g. , bouchard a. , blyth s. , van der heyden k. , prizkal n. , 2009 , in conference proceedings of the `` panoramic radio astronomy : wide - field 1 - 2 ghz research on galaxy evolution '' , june 02 - 05 , 2009 groningen , the netherlands s. , miller s. h. , penner k. , skelton r. e. , conselice c. j. , somerville r. s. , bell e. f. , zheng x. z. , rix h. , robaina a. r. , barazza f. d. , barden m. , borch a. , beckwith s. v. w. , caldwell j. a. r. , peng c. y. , heymans c. , mcintosh d. h. , huler b. , jahnke k. , meisenheimer k. , sanchez s. f. , wisotzki l. , wolf c. , papovich c. , 2009 , , 697 , 1971 s. , bailes m. , bartel n. , baugh c. , bietenholz m. , blake c. , braun r. , brown j. , chatterjee s. , darling j. , deller a. , dodson r. , edwards p. g. , ekers r. , ellingsen s. , feain i. , gaensler b. m. , haverkorn m. , hobbs g. , hopkins a. , jackson c. , james c. , joncas g. , kaspi v. , kilborn v. , koribalski b. , kothes r. , landecker t. l. , lenc e. , lovell j. , macquart j .- , manchester r. , matthews d. , mcclure - griffiths n. m. , norris r. , pen u .- phillips c. , power c. , protheroe r. , sadler e. , schmidt b. , stairs i. , staveley - smith l. , stil j. , taylor r. , tingay s. , tzioumis a. , walker m. , wall j. , wolleben m. , 2007 , publications of the astronomical society of australia , 24 , 174 s. , feain i. j. , gupta n. , 2009 , in astronomical society of the pacific conference series , vol . 407 , astronomical society of the pacific conference series , d. j. saikia , d. a. green , y. gupta , & t. venturi , ed . , pp . 446+ s. , taylor r. , bailes m. , bartel n. , baugh c. , bietenholz m. , blake c. , braun r. , brown j. , chatterjee s. , darling j. , deller a. , dodson r. , edwards p. , ekers r. , ellingsen s. , feain i. , gaensler b. , haverkorn m. , hobbs g. , hopkins a. , jackson c. , james c. , joncas g. , kaspi v. , kilborn v. , koribalski b. , kothes r. , landecker t. , lenc a. , lovell j. , macquart j .- , manchester r. , matthews d. , mcclure - griffiths n. , norris r. , pen u .- phillips c. , power c. , protheroe r. , sadler e. , schmidt b. , stairs i. , staveley - smith l. , stil j. , tingay s. , tzioumis a. , walker m. , wall j. , wolleben m. , 2008 , experimental astronomy , 22 , 151 j. m. , davis m. , faber s. m. , guhathakurta p. , gwyn s. , huang j. , koo d. c. , le floch e. , lin l. , newman j. , noeske k. , papovich c. , willmer c. n. a. , coil a. , conselice c. j. , cooper m. , hopkins a. m. , metevier a. , primack j. , rieke g. , weiner b. j. , 2008 , , 672 , 177 m. a. w. , oosterloo t. a. , van cappellen w. a. , bakker l. , ivashina m. v. , van der hulst j. m. , 2008 , in american institute of physics conference series , vol . 1035 , the evolution of galaxies through the neutral hydrogen window , r. minchin & e. momjian , ed . , pp . 265271 h. , dickinson m. , stern d. , eisenhardt p. r. m. , chary r .- , giavalisco m. , ferguson h. c. , casertano s. , conselice c. j. , papovich c. , reach w. t. , grogin n. , moustakas l. a. , ouchi m. , 2005 , , 634 , 109	we explore the quantified morphology of atomic hydrogen ( ) disks in the virgo cluster . these galaxies display a wealth of phenomena in their morphology , e.g. , tails , truncation and warps . these morphological disturbances are related to the ram - pressure stripping and tidal interaction that galaxies undergo in this dense cluster environment . to quantify the morphological transformation of the disks , we compute the morphological parameters of concentration - asymmetry - smoothness , gini and 20 and our own for 51 galaxies in 48 column density maps from the vla imaging of virgo spirals in atomic gas ( viva ) project . some morphological phenomena can be identified in this space of relatively low resolution data . truncation of the disk can be cleanly identified via the concentration parameter ( c@xmath01 ) and concentration can also be used to identify deficient disks ( 1@xmath0c@xmath05 ) . tidal interaction is typically identified using combinations of these morphological parameters , applied to ( optical ) images of galaxies . we find that some selection criteria ( gini-@xmath1 , asymmetry , and a modified concentration-@xmath1 ) are still applicable for the coarse ( @xmath215@xmath3 fwhm ) viva data . we note that asymmetry is strongly affected by the choice for the center of these galaxies . the phenomena of tidal tails can be reasonably well identified using the gini-20 criterion ( 60 % of galaxies with tails identified but with as many contaminants ) . ram - pressure does move disks into and out of most of our interaction criteria : the ram - pressure sequence identified by vollmer et al . ( 2009 ) tracks into and out of some of these criteria ( asymmetry based and the gini-@xmath1 selections , but not the concentration-20 or the based ones ) . therefore , future searches for interaction using morphologies should take ram - pressure into account as a mechanism to disturb disks enough to make them appear as gravitationally interacting . one mechanism would be to remove all the deficient ( c@xmath05 ) disks from the sample , as these have undergone more than one removal mechanism . [ firstpage ]
collisionless shocks are widely believed to be the primary acceleration mechanism giving rise to the ubiquitous existence of energetic particles in space . the theory of diffusive shock acceleration ( dsa ) was proposed some @xmath0 years ago @xcite and is currently believed to be the most important mechanism for a variety of astrophysical environments , for example , interplanetary shocks , the heliospheric termination shock , and shocks associated with supernova remnants . the theory predicts a universal power - law energy flux spectrum @xmath1 for strong shocks with a density compression ratio of 4 . the physical mechanism by which particles are accelerated from thermal energies to much higher energies where dsa is presumed to be applicable ( the injection problem ) has received some recent attentions but has not reached a common consensus explanation . many acceleration theories , for example , shock drift acceleration ( sda ) ( see reviews , * ? ? ? * ; * ? ? ? * ) and shock surfing acceleration @xcite have been proposed . the acceleration of low energy protons in the shocks containing large - scale pre - existing magnetic fluctuations is very efficient @xcite , which suggests that there may not be an injection problem . it is generally thought that , pre - accelerated particles will interact resonantly with magnetic turbulence which results in isotropization and diffusion . many previous works considering magnetic turbulence focused on the ions ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , the acceleration of electrons is less well understood since for electrons whose gyroradii are very small , the cyclotron resonance condition is not easily satisfied thus they can not interact resonantly with large - scale ambient turbulence on ion - scale . while the scattering provided by whistler waves @xcite is one possibility , @xcite proposed an attractive solution to the injection problem that does not require pitch - angle scattering , i.e. , conserving the first adiabatic invariant . the idea is that the low - rigidity particles , especially electrons , can move rapidly along meandering magnetic field lines and thus travel back and forth between shock front . the particles gain energy from the difference between upstream and downstream flow velocities . energetic electrons are often observed to be associated with collisionless shocks . accelerated electrons are thought to produce type ii radio bursts in the solar corona and interplanetary space . @xcite reported _ isee _ spacecraft measurements of upstream electrons ( @xmath2 kev ) of the earth s bow shock that originate from a thin region close to the point of tangency between interplanetary magnetic field lines and the shock surface . @xcite showed observations of energetic electrons associated with interplanetary shocks showing `` spike - like '' flux enhancements for energies @xmath3 kev . the spike events were observed at quasi - perpendicular shocks with @xmath4 , where @xmath5 is the angle between upstream magnetic field and shock normal . some shock crossings had no enhancements of energetic electrons which were reported to be associated with low shock speeds and small @xmath5 . @xcite presented data which shows energetic electrons are accelerated close to shock front . they also showed some accelerated electrons can escape far upstream of quasi - perpendicular interplanetary shocks . the clear evidence of electron acceleration by dsa is rare , but a recent example was discussed by @xcite showing the importance of whistler waves . in order to explain the energization of electrons within the shock layer , @xcite and @xcite developed analytic models for electron acceleration from thermal energies by adiabatic reflection by a quasi - perpendicular shock . this is known as fast - fermi acceleration . this theory describes a scatter - free electron acceleration process in a planar , time - steady shock . it obtains a qualitative agreement with observations at earth s bow shock in terms of the loss - cone pitch - angle distribution and energy range of accelerated electrons . @xcite used the combination of electron test particle simulation and 1-d hybrid simulation and verified wu s basic conclusions . the main energy source of fast fermi acceleration comes from the @xmath6 electric field which is the same as sda @xcite . it can also be demonstrated that fast - fermi acceleration and sda are the same process in two different frames of reference @xcite . thus one would expect electrons to drift in the direction perpendicular to the flow and magnetic field . for a single reflection , the fraction and energies of accelerated particles are limited ( e.g. , * ? ? ? @xcite proposed the basic outline for type ii solar radio bursts in which energetic electrons are accelerated through sda . it is expected that multiple reflections are required in order to explain herringbone structures in type ii bursts , where the electrons are accelerated to a fraction of the speed of light . more recently , @xcite studied electron acceleration in 2-d quasi - perpendicular shocks using test - particle simulations and self - consistent hybrid simulations . he focused on the effect of the rippling of the shock front on particle acceleration in highly oblique shocks with @xmath7 . the ripples , in this case , were produced by instabilities along the shock front . burgess found that the acceleration of electron by sda can be more efficient for a rippled shock . the shape of the resulting energy spectra has a flat plateau from the initial release energies to energies several times higher than this . above the flat plateau the spectra drop off steeply as @xmath5 become smaller . in this paper , we use test particle simulations combined with @xmath8-d hybrid simulations that include pre - existing large - scale magnetic field turbulence @xcite to study the shock acceleration of electrons . in addition to the effect of the large - scale turbulence , the shock microphysics occuring on ion length and time scales is also included . moreover , the shock front is rippled and distorted in response to the turbulence , which is also included in our model . in @xmath9 2 we describe the numerical method we used to combine the fields from the hybrid simulation and test - particle simulation to obtain the electron distribution . @xmath9 3 gives the main results of our simulation . in @xmath9 4 we summarize the main conclusions and discuss the implication of our work . investigating particle transport in the vicinity of a collisionless shock requires a spatial scale large enough for particles to propagate back and forth across the shock , and a spatial resolution small enough to include the detailed physics for particle scattering and shock microstructure . we implement a combination of a @xmath8-d hybrid simulation to model the fields and plasma flow and a test particle simulation to follow the orbits of a large number of energetic electrons . in the first step , we employ a two - dimensional hybrid simulation similar to previous work @xcite that includes pre - existing large - scale turbulence . in the hybrid simulation ( e.g. , * ? ? ? * ) , the ions are treated fully kinetically and thermal ( i.e. , non - energetic ) electrons are treated as a massless fluid . this approach is well suited to resolve ion - scale plasma physics which is critical to describe supercritical collisionless shocks . in this study , we consider a two - dimensional cartesian grid in the @xmath10 plane . all the physical vector quantities have components in three directions , but depend spatially only on these two variables . a shock is produced by using the so - called piston method ( for a discussion , see * ? ? ? * ) , in which the plasma is injected continuously from one end ( @xmath11 , in our case ) of the simulation box , and reflected elastically at the other end ( @xmath12 ) . this boundary is also assumed to be a perfectly conducting barrier . the pileup of density and magnetic field creates a shock propagating in the @xmath13 direction . to include the large - scale magnetic fluctuations , a random magnetic field is superposed on a mean field at the beginning of the simulation and is also injected continuously at the @xmath11 boundary during the simulation . the simplified one - dimensional fluctuations have the form @xmath14 , where @xmath15 is the averaged upstream magnetic field . the fluctuating component contains an equal mixture of right- and left - hand circularly polarized , forward and backward parallel - propagating plane alfven waves . the amplitude of the fluctuations is determined from a kolmogorov power spectrum : @xmath16 in which @xmath17 is the coherence scale of the fluctuations , see @xcite for more details . for the simulations presented in this study , we take @xmath18 , which is the size of simulation box in z direction . note that in addition to magnetic fluctuations , there are also velocity perturbations with @xmath19 ( alfven waves ) . for most of the parts in the paper , we consider a turbulence variance @xmath20 , where @xmath21 and @xmath22 are the magnitude of velocity perturbation and upstream alfven speed , respectively . we also discuss the effect of different values of turbulence variances . the size of the simulation box @xmath23 , where @xmath24 is the ion inertial length . the mach number of the flow in the simulation frame is @xmath25 , the averaged mach number in the shock frame is about @xmath26 . most of the results presented here are for averaged shock normal angle @xmath27 , but we also simulate the cases for @xmath28 and @xmath29 to examine the dependence of the acceleration efficiency on shock normal angle . the other important simulation parameters include electron and ion plasma beta @xmath30 and @xmath31 , respectively , grid sizes @xmath32 , time step @xmath33 , the ratio between light speed and upstream alfven speed @xmath34 , and the anomalous resistivity is taken to be @xmath35 . the initial spatially uniform thermal ion distribution was generated using @xmath36 particles per cell . different from previous studies , the consideration of large - scale magnetic fluctuations enables us consider the effect of pre - existing magnetic turbulence on electron acceleration , which has been shown to be important for low - energy ion acceleration @xcite since particle transport normal to the mean field is enhanced . however , the particle transport in full @xmath37-d turbulence can not be properly treated in a self - consistent way using available computation . as demonstrated by previous works @xcite , in the model with at least one ignorable coordinate , the center of gyration of particles is confined to within one gyroradius of the original magnetic field line . the test - electrons can still move normal to the mean field in our model because of the field - line random walk . in the second part of our calculation we integrate the full motion equation of an ensemble of test - particle electrons in the electric and magnetic fields obtained in the hybrid simulations ( see figure @xmath38 ) . this part of the calculation is done separately from the main hybrid simulation as a post processing phase . we assume non - relativistic motion which is reasonable because the highest energy electrons obtained in our study are still non - relativistic . as noted by @xcite , high - order interpolation of fields is required to ensure numerical accuracy and avoid artificial scattering in calculating electron trajectories . in this work we use second - order spatial interpolation and linear temporal interpolation , which ensure the smooth variations of the electromagnetic fields . we release a shell distribution of electrons with energy of @xmath39 ev , which corresponds to an electron velocity @xmath40 in the upstream frame , where @xmath41 is upstream bulk velocity in the shock frame and @xmath42 is the thermal velocity of fluid electrons considered in the hybrid simulations , respectively . this energy is typical for the halo component of electron velocity distributions observed in solar wind . the test - particle electrons are released uniformly upstream at @xmath43 when the shock has fully formed and is far from the boundaries . the numerical technique used to integrate electron trajectories is the so - called bulirsh - stoer method , which is described in detail by @xcite . it is highly accurate and conserves energy well . it is fast when fields are smooth compared with the electron gyroradius . the algorithm uses an adjustable time - step method based on the evaluation of the local truncation error . the time step is allowed to vary between @xmath44 and @xmath45 , where @xmath46 is the electron gyrofrequency . the ratio @xmath47 is taken to be the realistic value @xmath48 . the total number of electrons in the simulation is @xmath49 . the electrons which reach the left or right boundary are assumed to escape from the shock region and are removed from simulation . the boundary condition in the @xmath50 direction is taken to be periodic . the readers are referred to @xcite for more details on the numerical methods . magnetic field turbulence has already been proved to have key effect on the particle acceleration in collisionless shocks . unfortunately , solving the whole problem in three - dimensional space and resolving magnetic turbulence from coherence scale to electron scale are still limited by available computation in the near future . this limitation motivates us to solve these problems approximately . we also note that in our model the electron test particle simulation is not self - consistent since the hybrid simulation does not include the electron scale physics . the electron scale shock structure , which may be important is neglected here . figure @xmath38 shows a snapshot of the @xmath50 component of the magnetic field , @xmath51 , at time @xmath52 in a gray - scale representation , overlayed by a two - dimensional magnetic vector field . at this time , the shock is fully developed . in this case , the angle between the average magnetic field direction and shock normal , @xmath53 is @xmath54 . the position of the shock front is clearly seen from the boundary of the magnetic field jump . the shock is moving in the @xmath13 direction at a speed dependent on @xmath50 , which is about 1.6 @xmath22 on average . because of the effect of large - scale turbulence with the shock , the shock surface become irregular on a variety of spatial scales from small - scale ripples , which could be due to ion - scale plasma instabilities @xcite , to large - scale structure caused by the interaction between the shock and upstream turbulence @xcite . the upstream magnetic field is compressed and distorted as it passes through the shock into the downstream region . we note that the rippling of the shock and varying upstream magnetic field leads to a varying local shock normal angle along the shock front . as we will discuss later , the irregular shock surface and magnetic field geometry will efficiently accelerate electrons and produce a number of features similar to observations , such as the electron foreshock and spike - like intensity increases at the shock front . the meandering of field lines close to the shock surface helps to trap the electrons at the shock , leading to efficient acceleration . the shock ripples also contribute to the acceleration by mirroring electrons between them . figure @xmath8 shows a color - coded representation of the number of energetic electrons with energies higher than @xmath55 times ( i.e. , @xmath38 kev ) the initial ( at release ) energy at three different times ( a ) @xmath56 , ( b ) @xmath57 , ( c ) @xmath58 , respectively . it is found that after the initial release , a fraction of the electrons are reflected and accelerated at the shock front , and then travel upstream along the turbulent magnetic field lines . these accelerated electrons are then taken back to the shock by the field line meandering , which provides even further acceleration . the number of energetic electrons close to the shock surface is highly irregular because the acceleration efficiency varies along the shock front depending on the local shock normal angle @xcite . most of the electrons are concentrated near the shock front since the global magnetic field is mostly perpendicular to the shock normal . as the field lines convect through shock , the electrons eventually are taken downstream . since the electrons are tied to individual field lines in 2-d magnetic field , once the electrons are no longer capable of crossing the shock , there will be no additional significant acceleration . at this point , once all electrons are downstream , the energy spectrum no longer changes with time . examination of the trajectories of some electrons shows that the rippling of the shock front also contributes to the acceleration by mirroring electrons between the ripples , as illustrated in figure @xmath37 . in this figure , the top left plot displays the trajectory of a representative electron in the @xmath10 plane , overlapped with the 2-d gray - scale representation of @xmath59 at @xmath60 , the gray scale is the same as in figure @xmath38 . the upper right plot shows the position of this electron ( in @xmath61 ) as a function of time . the electron bounces back and forth between the ripples for several times . for example , the reflections are labeled @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 , and @xmath67 . the energy change as a function of position , @xmath61 , corresponding to these reflections is shown in the bottom left panel . we find that there are jumps in energy at each of the reflections . the panel on the bottom right shows the electron energy as a function of time which also illustrates the features of multiple accelerations related to multiple reflections . the trajectory analysis shows the electron will be mirrored between the ripples for a couple of times and by this get accelerated multiple times . note that the shock does not move much during the time scale of this trajectory . we now consider the effect of varying the angle between the mean magnetic field and shock - normal . shown in figure @xmath68 are the resulting energy spectra for three different mean shock - normal angles ( @xmath69 and @xmath54 , respectively ) at the end of simulations ( @xmath70 ) . it is found that for @xmath27 , the electrons can readily be accelerated to up to @xmath71 times the initial energy within @xmath72 . the spectrum is flat between about @xmath73 kev to @xmath74 kev . this shape is similar to the `` plateau '' structure discussed by @xcite . above @xmath38 kev , the spectrum falls off with energy as a slope index about @xmath75 . it can be found that both the number fraction and highest energy of accelerated particles decreases as @xmath53 decreases . we have also tried different value of initial energies ( not shown ) , and find that the acceleration efficiency decreases for electrons with higher initial energies , which is similar to the results of @xcite . the effect of different strengthes of magnetic turbulence is examined in figure @xmath76 . we compare three cases with different turbulence variances @xmath77 , @xmath78 , and @xmath79 , respectively . at the end of simulations , the final energy spectra are similar at low energies , with significant variations in the spectra only at energies larger than @xmath8 kev . it is found that the energy spectrum is hardened at high energies when the turbulence variance is largest , which indicates the large - scale turbulence is more important for accelerating electrons to high energies . we argue that collisionless shocks which move through magnetic turbulence with significant power leads to efficient electron acceleration to high energies since the motion normal to the shock front is enhanced . the reason is that the meandering of field lines is enhanced , which allows the electrons have a better chance to travel though the shock multiple times . we note that the spatial distribution of energetic electrons is determined not only by the ripples in the shock front , but also by the global topology of the magnetic field lines . an example is shown in figure @xmath80 , which shows the profiles of the number of energetic electrons at @xmath81 as a function of @xmath61 , for the case of @xmath82 . the black solid line is the profile at @xmath83 , and the red dash line shows the profile at @xmath84 . the corresponding position of the shock front at each of these values of @xmath50 are represented using dot lines . at @xmath85 , it is observed that the energetic electrons travel far upstream up to about @xmath86 . however , the profile at @xmath87 shows no significant upstream energetic electron flux . the upstream energetic electron profiles show irregular features similar to in - situ observations reported by @xcite ( in figure 10 ) . the irregular features are controlled by the global topology of the large - scale turbulent magnetic field lines , along which the accelerated electrons could travel far upstream . additionally , energetic electron profiles in @xmath61 direction generally show `` spike - like '' structure close to the shock front , which is usually observed in interplanetary shocks and earth s bow shock . we note this feature is relatively stable within the simulation time once the upstream electron structure developed . we studied the acceleration of electrons at collisionless shocks by utilizing a combination of a 2-d hybrid simulation to obtain the shock structure and a test - particle simulation to determine the motion of electrons . the hybrid simulation provides realistic electric and magnetic fields within the transition layer of the shock that effect the motion of test - electrons , which is determined by solving the equation of motion . the interaction of the shock with pre - existing upstream fluctuations , and other nonlinear processes occuring in the hybrid simulation lead to a `` rippling '' of shock surface which also effects the transport of the electrons . we find that the electrons are efficiently accelerated by a nearly perpendicular shock . the turbulent magnetic field , leads to field - line meandering which allows the electrons to cross the shock front many times . the rippling of the shock front also contributes to the acceleration by mirroring electrons between the ripples . for the case that the averaged shock normal angle @xmath27 and turbulence variance @xmath88 , the electrons can readily be accelerated to up to @xmath71 times the initial energy . the resulting spectrum is flat between about @xmath73 kev to @xmath74 kev . at higher energies , the spectrum falls off with energy like a power law with a spectral slope of about @xmath75 . this acceleration process is more efficient at perpendicular shocks . as @xmath53 decreases from @xmath54 , both the number fraction and highest achievable energy of accelerated particles decreases . based on our calculations , we conclude that perpendicular shocks are the most important for the acceleration of electrons . the current study is helpful in understanding the injection problem for electron acceleration by collisionless shocks . it is also found that different value of magnetic turbulence variances strongly affects the maximum energy attainable . the case with larger turbulence variance has a flatter energy spectrum than the case of smaller turbulence variance , which suggests the enhanced motion of electrons normal to the shock front , due to enhanced field - line random walk , is of importance for the acceleration of electrons to high energies . in addition , we also found that the energetic electron density upstream and downstream of collisionless shocks show filamentary structures ( figure 2 ) . this could help explain electron spike - like events observed upstream and downstream of terrestrial and interplanetary shocks @xcite . observation by voyager @xmath38 at the termination shock and in the heliosheath also show the evidence of electron spike - like enhancements at the shock front @xcite . the upstream spatial distribution of energetic electrons shows irregular features which depend on both the irregularity in the shock surface and the global topology of magnetic field lines . at first the electrons are accelerated and reflected at the shock front , and then they travel upstream along the magnetic field lines . the electrons could be taken far upstream by field line random walk . this result can possibly lead to an interpretation to the complex electron foreshock events recently observed to be associated with interplanetary shocks @xcite . @xcite and @xcite proposed the complex upstream electron events are resulted from large - scale irregularities in shock surface . in this paper we have demonstrated that the upstream electron flux may be controlled by both an irregular shock surface and large - scale meandering magnetic field lines . the authors would like to thank dr . d. burgess for sharing the details of numerical methods and dr . j. r. jokipii for valuable discussions . this work was supported in part by nsf under grant atm0447354 and by nasa under grant nnx07ah19 g . direction @xmath51 represented in gray - scale at time @xmath89 , where @xmath90 is the averaged upstream magnetic field strength . a two - dimensional vector field is also overlapped , which indicates the direction of magnetic field . the shock surface is shown to be rippled and irregular in different scales . ] plane , overlapped with contour of @xmath59 magnetic field where the gray - scale is the same as that in figure 1 ; the top right panel shows the position of the electron in @xmath61 coordinate as a function of time ; the bottom left panel illustrates the energy of the representative electron @xmath96 as a function of @xmath61 ; the bottom right panel shows the dependence of electron energy @xmath96 on time . ] for different averaged shock normal angle . the red solid line is in the case that the shock angle @xmath82 , the blue dot dashed line and the black dashed line are in the cases that @xmath97 and @xmath98 , respectively . ] for averaged perpendicular shock with different turbulence variances . the black dashed line , red solid line , and blue dot dashed line are in the cases that @xmath77 , @xmath78 , and @xmath79 , respectively . ]	we study the physics of electron acceleration at collisionless shocks that move through a plasma containing large - scale magnetic fluctuations . we numerically integrate the trajectories of a large number of electrons , which are treated as test particles moving in the time dependent electric and magnetic fields determined from 2-d hybrid simulations ( kinetic ions , fluid electron ) . the large - scale magnetic fluctuations effect the electrons in a number of ways and lead to efficient and rapid energization at the shock front . since the electrons mainly follow along magnetic lines of force , the large - scale braiding of field lines in space allows the fast - moving electrons to cross the shock front several times , leading to efficient acceleration . ripples in the shock front occuring at various scales will also contribute to the acceleration by mirroring the electrons . our calculation shows that this process favors electron acceleration at perpendicular shocks . the current study is also helpful in understanding the injection problem for electron acceleration by collisionless shocks . it is also shown that the spatial distribution of energetic electrons is similar to in - situ observations ( e.g. , * ? ? ? * ; * ? ? ? * ) . the process may be important to our understanding of energetic electrons in planetary bow shocks and interplanetary shocks , and explaining herringbone structures seen in some type ii solar radio bursts .
the hubbard model @xcite is one of the generic models in many particle physics . due to difficulties with the analytic solution of the hubbard model in two dimensions , this model is intensively studied with various numerical algorithms , e.g. exact diagonalization @xcite , @xcite , @xcite @xcite , stochastic diagonalization @xcite , @xcite and quantum monte carlo algorithms @xcite , @xcite , @xcite , @xcite . the single band hubbard model with additional next nearest neighbor hopping is given in real the space by : @xmath2 the sum @xmath3 is over the nearest neighbors and @xmath4 is the sum over the next nearest neighbors . @xmath5 is the creation operator for an electron with spin @xmath6 on site @xmath7 and @xmath8 is the corresponding number operator . throughout this article we take @xmath9 as energy unit . in the momentum space this hamiltonian reads as : @xmath10 with @xmath11 the usual hubbard - model ( @xmath12 ) has a van hove singularity in the density of states at half filling in the noninteracting case ( @xmath13 ) . it is possible to move this van hove singularity to any electron filling by extending the hubbard model by an additional next nearest neighbor hopping @xmath1-term in the kinetic energy . the @xmath0-hubbard model @xcite shows superconductivity for repulsive interactions @xmath14 with d@xmath15-symmetry @xcite and for attractive interaction with on site s - symmetry @xcite , @xcite , @xcite . as a measure for the superconductivity we calculate the reduced two particle density matrix according to the concept of yang @xcite . from this two particle density matrix we calculate the two particle correlation functions for different symmetries and additionally we calculate the vertex correlation function @xcite . the van hove scenario predicts an increase of @xmath16 for fillings close to a van hove singularity @xcite . we study the influence of the van hove singularity on the superconducting correlation functions by modifying the @xmath1-hopping parameter for fixed fillings . here we use the lanczos - algorithm @xcite , @xcite as exact diagonalization technique to determine the ground state and from there ground state properties of the @xmath0-hubbard model . the basic limitations on the calculations of large system sizes with the lanczos - algorithm is the huge memory consumption of this method . but to our surprise we found that even for hubbard systems of size @xmath17 , which the lanczos method is capable of handling , the cpu consumption of the simulations was substantial and made a detailed scan of the parameter space given by interaction strength @xmath14 , filling @xmath18 and next nearest neighbor hopping @xmath1 almost impossible . we therefore implemented the lanczos - method on ibm sp2 parallel computer with mpi ( message passing interface ) for the communication between the processes to speed up the calculations . before turning our attention to the parallel techniques we outline the basic concept of the lanczos method @xcite . in the case of the exact diagonalization the hamiltonian @xmath19 of the system is written in matrix or heisenberg representation . one chooses an orthonormal single particle basis , to represent the many - particle states . for the @xmath0-hubbard model we use the momentum - space representation . each many particle basis state @xmath20 is a product of the spin up @xmath21 and the spin down @xmath22 component : @xmath23 and spin @xmath6 . @xmath24 is the vacuum state . each many particle state @xmath25 can be represented by means of the basis states @xmath26 and coefficients @xmath27 : @xmath28 the hilbert space of a system consisting of a lattice of @xmath29 sites and @xmath30 electrons with spin up and @xmath31 electrons with spin down has the dimension @xmath32 as the size of the hilbert space also determines the size of the computer memory , that is used in the calculation , one tries to reduce the hilbert space . the usual way to restrict the size of the hilbert space @xmath33 is to apply symmetries of the lattice and the hamiltonian . the translation invariance is a symmetry easily implemented when solving the @xmath0-hubbard model in momentum space representation with the lanczos algorithm . the hilbert space decomposes into subspaces @xmath34 containing only basis states @xmath20 , which have the same total momentum @xmath35 : @xmath36 where @xmath37 and @xmath38 are the momenta of the creation operators @xmath39 resp . @xmath40 in eq . [ eqphi ] . we denote the number of states of particles with the same spin with the total momentum @xmath41 as @xmath42 . in a @xmath17 system with @xmath43 the number of states @xmath44 varies for different momentum between @xmath45 and @xmath46 . the states @xmath47 must have the total momentum @xmath48 , so that the product state latexmath:[$\|\phi_{i_\uparrow,\uparrow } \rangle \otimes @xmath34 . altogether the subspace @xmath34 has the size @xmath50 for various fillings @xmath51 table [ tabstates ] gives an overview . when storing the coefficients @xmath27 with 8 byte floating point numbers the memory demand for one state is @xmath52 mbyte for a lattice with 16 sites and @xmath53 electrons . even if one uses the translation symmetry it remains a demand of @xmath54 mbyte for each state . in the standard lanczos algorithm @xcite , @xcite it is necessary to store three many particle states @xmath55 to calculate the ground state and the ground state energy . we implemented the lanczos - iteration @xcite : @xmath56 with the coefficients @xmath57 ( diagonal ) and @xmath58 ( offdiagonal ) of the tridiagonal matrix @xmath59 and the lanczos - vectors @xmath60 . from this tridiagonal matrix @xmath59 we calculate the eigenvalues of @xmath19 . in the lanczos - scheme it is not necessary to transform the matrix @xmath19 . therefore it is even not necessary to store the matrix elements @xmath61 ; they are only calculated , when they are needed for the further evaluation of the lanczos - iteration ( eq . [ eqlanczos ] ) . in this section we concentrate our effort on how to speed up the simulations with the lanczos method . the determination of the ground state properties of a hamiltonian with the lanczos algorithm consists of two main parts concerning the consumption of cpu - time . first the ground state energy @xmath62 and the ground state @xmath63 are calculated . the second main part is the determination of the two particle density matrix @xcite : @xmath64 as the main observable of interest . to handle an arbitrary state @xmath55 it is necessary to know for each basis state @xmath26 the indices @xmath37 and @xmath65 of the creation operators in eq . [ eqphi ] and the weights @xmath27 . a simple possibility for such an algorithm is the bitcoding or bitrepresentation of the basis states . here the momenta @xmath37 are labeled from @xmath66 to @xmath29 and the momenta @xmath65 from @xmath67 to @xmath68 . then the states @xmath26 are expressed by an one dimensional array of bits , which are one for occupied sites , and otherwise the bits are zero . if one interprets this array as binary representation of an integer number one has an algorithm to assign each basis state @xmath69 an index @xmath70 . as example we take a lattice with 4 sites and each two electrons with spin up and down : @xmath71 for any many particle state only the coefficients @xmath72 need to be stored . but the bitrepresentation has the great disadvantage , that a huge amount of memory is wasted , because the bitrepresentation of many integer numbers @xmath70 does not correspond to a valid basis state . for a system with @xmath29 latticepoints in an array of the length @xmath73 one only stores @xmath74 numbers . for the above example @xmath75 and @xmath53 there is @xmath76 . therefore it is desirable to use another algorithm ( @xmath77 ) . one example is the hashing algorithm @xcite . we developed a new algorithm . though we only present results for the momentum space representation , this algorithm can also be implemented very efficiently for the exact diagonalization in real space @xcite . first we are numbering the momenta from 1 to @xmath29 ( @xmath78 ) . second we define @xmath79 to fix the sign . the state with the number @xmath66 is @xmath80 in @xmath81 the electron @xmath82 moves from @xmath83 to @xmath84 . in @xmath85 this `` last '' electron has reached the latticepoint @xmath29 . next in @xmath86 the two creation operators with the highest index are increased by one , @xmath87 then the `` last '' creation operator moves . these are the states with the numbers @xmath88 to @xmath89 . next the `` last '' two electrons go to the sites @xmath84 and @xmath90 . if both electrons have reached the final two lattice sites ( @xmath91 , @xmath29 ) three electrons move in the same manner through the lattice . one gets the number @xmath92 of the state @xmath93 with @xmath94 where @xmath95 is the position of the electron @xmath96 in the lattice . using the translational invariance of the hamiltonian in the k - space representation means , that we keep the total momentum @xmath35 of the basis states @xmath97 fixed . in this case one can only choose the `` spin - up '' part of the basis state free and take such a `` spin - down '' state , that eq . [ eqk ] is full filled . this means , we must use another convention to label the states . we generate all states @xmath98 in the sequence as described above and calculate the momentum . for each momentum we count independently the indices . the coefficients @xmath27 are now labeled in following way : first we combine with state number 1 , momentum 1 and spin up with all possible states with spin down for a given @xmath35 . then we do the same with state number 2 , momentum 1 and spin up . next we switch to state 1 with momentum 2 and spin up and so on . in the parallel algorithm each of the @xmath99 processes stores the coefficients @xmath27 of @xmath100 basis states . @xmath27 is stored on process @xmath101 . in the lanczos method it is necessary to perform a matrix - vector multiplication between the matrix @xmath19 and a lanczos - vector @xmath102 . in the parallel algorithm this is carried out in the following way : for @xmath103 : * each process @xmath104 calculates the index numbers @xmath70 of the basis states @xmath105 , the multiplication results for the @xmath7-th state and the process , on which the states @xmath70 are stored . * each process exchanges the multiplication results and the numbers @xmath70 of the basis states , which are not stored on the process , with process @xmath106 . to calculate the two - particle density matrix @xmath107 in an efficient way , one only calculates the elements @xmath108 which are nonzero . that means one takes a basis state @xmath109 , one of the @xmath110 possible combinations of @xmath111 , @xmath112 , @xmath113 and @xmath114 and applies @xmath115 to @xmath116 , afterwards one calculates the index number of this transformed basis state @xmath117 . each process stores the complete matrix @xmath107 . this matrix takes for example in the @xmath17 system @xmath118 byte or @xmath119 kbyte of memory . in this algorithm one has to calculate @xmath120 expectation values , compared to @xmath121 expectation values that would be calculated if one took all combinations @xmath97 and @xmath122 into account . in a @xmath123 system with @xmath53 electrons @xmath124 and the amount of saved cpu - time is significant . but in a @xmath17 system with only @xmath125 electrons @xmath126 and this algorithm is slower . to calculate @xmath107 we use a similar way for the exchange of the weights and numbers as for matrix - vector multiplication ( see section [ sectionmatmult ] ) . at the end the values of the arrays @xmath107 of all processes are summed on one process . first we take a look on the dependence of the cpu time of one lanczos iteration ( eq.[eqlanczos ] ) for a different number of processes . as example we use a @xmath123 lattice with three different numbers of electrons ( fig . [ cpuiter ] ) . the small decay from 1 to 2 processes is due to communication between processes which is only necessary for more than one process . then one sees as expected a decrease of computation time . as expected this decrease vanishes for an increasing number of processes , since the communication is growing with the number of processes . in figure [ cpucorr ] we examine the cpu - consumption for the two - particle density matrix for the same system size and fillings as in figure [ cpuiter ] . here the gain of time is much larger than for the calculation of the energy . in this case more computation is performed for the determination of one matrix element . for more than 2 processes the dependence of nodes and cpu - time is nearly linear . this can be understood , if we look on the numbering of the states . most of the states @xmath127 are stored on the same node and only for a fraction of these states the weights must be interchanged with an other process . summarizing the algorithm for this parallel implementation of the lanczos - iteration and the determination of the two - particle density matrix is a coarse grained algorithm and therefore achieves a good speed - up on a parallel computer like the ibm sp2 , with only some , but very powerful , processors . for very many processors the communication grows dramatically and no further speed - up is reachable . now we want to turn our attention from the technical points of view to physical properties of the @xmath0-hubbard model . especially we study the influence of the next nearest hopping parameter @xmath1 on the ground state energy and superconducting correlation functions in the ground state of a @xmath17 cluster . we focus to @xmath128 electrons which corresponds to a filling @xmath129 . this is a so called closed shell situation , which can be also handled with the projector quantum monte carlo method @xcite . first we study the ground state energy @xmath62 in the attractive @xmath0-hubbard model ( fig . [ emin ] ) . for small and intermediate interaction strength ( @xmath130 ) there is a visible difference between the energy with @xmath12 and @xmath131 . this difference results from the changes in the structure of @xmath132 ( eq.[eqepsilon ] ) with @xmath1 . but for large interaction strength @xmath133 the influence of the kinetic part is vanishing . in this interaction regime the ground state energy is approximately linear with @xmath14 and approaches slowly the energy @xmath134 of the system without hopping . next we turn our attention to the superconducting correlation functions . in the concept of off diagonal long range order @xcite the largest eigenvalue and eigenvector of @xmath107 is calculated . the quantum monte carlo algorithms handle system sizes , where it is impossible to calculate the complete two - particle density matrix due to the memory consumptions @xcite . in order to compare the exact diagonalization results with the quantum monte carlo data we study two - particle correlation functions for certain symmetries @xcite , e.g. @xmath135 where the index s denotes the on site s - wave symmetry and d the @xmath136-wave symmetry . the factor @xmath137 gives the signs of the d - wave . ( @xmath138 in x- and @xmath139 in y - direction ) these full correlation functions have nonzero values even for a system with no interaction . responsible for this are the one - particle correlation functions @xmath140 which decay to zero with @xmath141 and thus do not really contribute to the long range behavior that signals superconductivity . to exclude the contribution of the one - particle correlation functions we define the vertex correlation function as @xmath142 in figure [ corrvgl ] we show @xmath143 and @xmath144 in dependence of the distance @xmath145 . for @xmath146 there is a visible difference only for @xmath147 . for @xmath148 the one - particle contributions are dominant . in this case it is important to study the vertex correlation function to get the `` superconducting '' correlations . but already for @xmath149 and @xmath150 the difference between full and vertex correlation function is less than 30% and it is less important to take @xmath151 in account . as a measure for the superconductivity in a system we show in fig . [ neguvertex ] the average of the vertex correlation function @xmath152 for a small interaction strength @xmath133 the increase of the correlation functions is small . between @xmath149 and @xmath153 there is a strong increase . finally at @xmath154 the curves flatten . for correlation functions the influence of the additional hopping @xmath1 remains important even if there is nearly no difference in the energies ( cp . [ emin ] and [ neguvertex ] , @xmath155 ) . therefore we study the influence of @xmath1 for the interaction @xmath156 ( fig . [ negutprime ] ) . the correlation functions have a broad maximum around @xmath157 . as it is commonly accepted @xcite the finite size gap has an influence on the superconducting correlation functions . the finite size gap @xmath158 in the energy dispersion @xmath132 ( eq . [ eqepsilon ] ) of the free system ( @xmath13 ) between the highest occupied state and the lowest unoccupied state is given in this case by : @xmath159 this means the finite size gap is becoming smaller with increasing @xmath160 . according to @xcite this will lead to an increase in the superconducting correlations . figure [ negutprime ] confirms this for @xmath161 . but at @xmath162 the correlations decrease again . therefore also the structure of the energy dispersion @xmath132 has an influence on the correlation functions . in figure [ negutprime ] the maximum of the correlation functions is not at @xmath163 , where the noninteracting system has a van hove singularity in the thermodynamic limet . there a maximum of @xmath16 is predicted by the van hove scenario @xcite . but for small system sizes one can not really speak of a van hove singularity . therefore the results of figure [ negutprime ] are not in contradiction to the van hove scenario . yet , the question remains , whether the observed increase of the correlation function results only from a vanishing finite size gap or is related to changes of the fermi surface . to clarify this point it will be necessary to calculate larger systems . next we study the influence of the @xmath1 hopping for repulsive interaction @xmath164 . quantum monte carlo calculations show a plateau for the d@xmath15 correlation function in the @xmath0-hubbard model @xcite , @xcite . figure [ posutprime ] shows the average correlation functions with @xmath136 symmetry . here the full correlation function is nearly independent of @xmath1 for @xmath165 . the vertex correlation function is much smaller than in the attractive case ( cp . [ negutprime ] and fig . [ posutprime ] ) . the increase of @xmath166 by a factor of about 10 between @xmath12 and @xmath167 is much larger than the increase of @xmath168 and @xmath169 in the attractive @xmath0-hubbard model and of @xmath170 in the repulsive model . in figure [ posulong ] the vertex correlation function with @xmath136 symmetry is plotted against the distance @xmath145 of the `` cooper pairs '' . for @xmath171 the vertex correlation function has , in contrast to larger @xmath1 , no longer a negative value at @xmath172 and is positive for all distances @xmath145 . this negative value at @xmath172 is also seen in the quantum monte carlo results for larger systems @xcite , @xcite . in the case of @xmath157 , where the gap @xmath158 gets zero , @xmath173 changes its shape completely and most values are negative and also the average is negative . in contrast to the attractive @xmath0-hubbard model , where the plateau is decreasing gradually , in the repulsive case one observes a complete break down of the plateau for @xmath174 . this means again , that the vertex correlation function depends strongly on the energy dispersion @xmath132 ( eq.[eqepsilon ] ) , which is transformed due to the @xmath1-hopping . as in the attractive @xmath0-hubbard model the maximum of the superconducting correlations is not at @xmath163 , where a van - hove singularity is in the non interacting infinite system . to clarify the origin of this behavior it is necessary to study larger systems . for a possible connection with the van - hove - scenario it is also necessary to calculate the density of states for this parameter regime we have presented an effective algorithm for the implementation of the exact diagonalization of the @xmath0-hubbard model in momentum - space respresentation . as method for the exact diagonalization we use the lanczos algorithm . we showed a detailed description of the parallel algorithm . the speed - up of the code is almost linear for the correlation functions and is increasing with increasing size of the hilbert space . the key point in our algorithm is a new method of labeling the states that is compact and in contrast to previous methods @xcite , @xcite also gives a consecutive order without any interruption . this makes the distribution on the different processes for parallelization a straightforward task . with the access to powerful modern parallel computers we were able to scan the parameter space of the @xmath0-hubbard model in more detail . the influence of the @xmath1-hopping to the ground state energy is vanishing with increasing interaction strength in the attractive @xmath0-hubbard model . this is in contradiction to the correlation functions with on site s - wave symmetry , where the influence of @xmath1 is remaining for all studied attractive interactions . the average full and vertex on site s - wave correlation functions have a broad maximum at @xmath175 , where the gap @xmath158 is vanishing , in the attractive @xmath0-hubbard model . in the repulsive @xmath0-hubbard model only the d@xmath15 vertex correlation functions show a strong increase with decreasing @xmath1 and gap @xmath158 . at @xmath175 and @xmath176 @xmath177 has a break down to a negative value . the origin of this behavior and a possible connection with the van - hove scenario for high @xmath16 superconductors is yet not clear . simulations for larger systems are necessary . we are grateful for the leibnitz rechenzentrum mnchen ( lrz ) for providing us a generous amount of cpu - time on their ibm sp2 parallel computer . werner fettes wants to thank the `` deutsche forschungs gemeinschaft '' ( dfg ) for the financial support .	we present a new parallel algorithm for the exact diagonalization of the @xmath0-hubbard model with the lanczos - method . by invoking a new scheme of labeling the states we were able to obtain a speedup of up to four on 16 nodes of an ibm sp2 for the calculation of the ground state energy and an almost linear speedup for the calculation of the correlation functions . using this algorithm we performed an extensive study of the influence of the next - nearest hopping parameter @xmath1 in the @xmath0-hubbard model on ground state energy and the superconducting correlation functions for both attractive and repulsive interaction .
in this paper we study the asymptotic properties related to connectivity of random geometric graphs where the underlying distribution of the vertices may not be uniform . a random geometric graph ( rgg ) consists of a set of vertices that are distributed in space independently , according to some common probability density function . the edge set of the graph consists of the set of all pairs of points that are within a specified cut - off distance . our point of departure from usual random geometric graphs is the specification of a cut - off function @xmath7 that determines the edge set . a directed edge exists , from a vertex located at @xmath4 to another vertex located at @xmath8 , provided the distance between @xmath4 and @xmath8 is less than @xmath9 . our motivation for the study of such graphs comes from applications in wireless networks . in models of wireless networks as rggs , the nodes are assumed to be communicating entities that are distributed randomly in space according to some underlying density . nodes are assumed to communicate effectively with other nodes that are within a cut - off distance , that is , proportional to the transmission power . hence , the transmission power has to be sufficiently large for the network to be connected . however , nodes that are within each other s transmission range interfere and thus can not transmit simultaneously . in order to maximize spatial reuse , that is , the simultaneous use of the medium by several nodes to communicate , the transmission power should be minimized . thus the asymptotic behavior of the critical radius of connectivity in a random geometric graph as the number of vertices becomes large is of considerable interest . often the nodes are assumed to be distributed in @xmath10^d$ ] according to a poisson point process of intensity @xmath1 . in this case , it is known that the critical connectivity radius scales as @xmath11 if the underlying density is nonuniform but bounded away from zero , then ( see penrose @xcite ) the asymptotic behavior of the largest nearest neighbor distance in the graph is @xmath12 where @xmath13 is the minimum of the density over its support . note that the asymptotics of the largest nearest neighbor distance is determined by the reciprocal of the minimum of the density , since it is in the vicinity of the minimum that the nodes are sparsely distributed . in many applications such as mobile ad - hoc networks and sensor networks , the distribution of the nodes may be far from uniform ( see , e.g. , foh et al . @xcite , santi @xcite ) . in the case of nonuniform distribution of nodes , it is not efficient from the point of view of maximizing spatial reuse , for all nodes to use the same cut - off radius . nodes near the mode of the density require a much smaller radius than those at locations where the density is small . further , the infimum of the density over its support could be zero ( @xmath14 ) . in such cases the asymptotics of the largest nearest neighbor distance or the connectivity threshold will be very different from that given above . one of the major objectives in a wireless sensor network is to maximize battery life , and hence it is important to minimize the energy expended in data transmission . these considerations leads us to rggs with location - dependent choice of radii . in many applications , it is assumed that the nodes know or can effectively estimate their location ( akkaya and younis @xcite , langendoen and reijers @xcite ) . we prescribe a formula for a critical location dependent cut - off radius depending on the intensity , so that almost surely the resulting graphs do not have isolated nodes eventually . a useful property of the graphs we construct is that the distribution of out - degree is independent of the location of the nodes . vertex degree distributions are important in designing algorithms for distributed computations over wireless networks where the performance worsens with increasing vertex degrees ( giridhar and kumar @xcite ) . we derive strong law bounds for the maximum and minimum vertex degrees . by considering a finer parametrization of the cut - off function , we show that the number of vertices with zero out - degree converges to a poisson distribution , under some conditions on the underlying density . we illustrate the result with some examples . a result of this nature for usual random geometric graphs with the uniform and exponentially decaying densities of nodes can be found in penrose @xcite , gupta and iyer @xcite , respectively . the solution to the connectivity problem for random geometric graphs for nonvanishing densities with compact support and dimensions @xmath15 can be found , for example , in chapter 13 of penrose @xcite . in one dimension , this problem is studied for densities in @xmath10 $ ] with polynomial rate of decay to zero in han and makowski @xcite , while in gupta , iyer and manjunath @xcite the density is assumed to be exponential or truncated exponential . in two dimensions , the asymptotic distribution for the critical connectivity threshold for a large class of densities , including elliptically contoured distributions , distributions with independent weibull - like marginals and distributions with parallel level curves is derived in hsing and rootzen @xcite . in dimensions @xmath15 , penrose @xcite obtains the asymptotic distribution for the connectivity threshold when the nodes are distributed according to a standard normal distribution . in this paper , we derive a sufficient condition for the rggs with location - dependent radius to be almost surely connected eventually . in summary , our primary motivation in proposing the study of graphs with location dependent radii is as follows . it is to enable the design and study of wireless and sensor networks that allow nonstandard distribution of nodes obtained by fitting densities to empirical data obtained from actual deployments . given such a density , each individual node can be programmed to choose a transmission radius depending on its location so that the network is connected with high probability . further , any change in the underlying distribution over time due to failures , re - deployments , etc . , can be easily accommodated by appropriately changing the transmission radii . as far as analyzing these graphs is concerned , the key features to contend with are that the edges are directed and that the cut - off radius is specified implicitly . in order to state our results we need some notation . let @xmath2 be a continuous probability density function with support @xmath16 . for any random variable @xmath17 with density @xmath2 , we denote it by @xmath18 . let the metric on @xmath3 be given by one of the @xmath19 norms @xmath20 , denoted by @xmath21 . let @xmath22 denote the volume of the unit ball in @xmath3 . we denote by @xmath23 the closure of the set @xmath24 . let @xmath25 be a sequence of i.i.d . points distributed according to @xmath2 . let @xmath26 be a nondecreasing sequence of poisson random variables with @xmath27 = n$ ] and define the sequence of sets @xmath28 note that @xmath29 is a poisson point process with intensity @xmath30 . for any @xmath31 and @xmath32 we denote by @xmath33 the open ball of radius @xmath34 centered at @xmath4 . for any borel set @xmath35 and any point process @xmath36 , @xmath37 $ ] represents the number of points of @xmath36 in @xmath24 , and define @xmath38 we now define the random geometric graphs of interest with location - dependent radii . [ defrgg ] for any @xmath39 , let @xmath29 be the set given by ( [ eqnpn ] ) . for any function @xmath40 , the random geometric graph @xmath41 is defined to be the graph with vertex set @xmath29 , and the directed edge set @xmath42 we will also consider an augmented version of the above random geometric graph which is obtained by making all the edges in @xmath41 bi - directional . [ defrgge ] the enhanced random geometric graph @xmath43 associated with the graph @xmath41 is defined to be the graph with vertex set @xmath29 and ( undirected ) edge set @xmath44 in the communication application described in the , the following procedure will give a graph whose edge set will _ contain _ the edges of the enhanced graph @xmath45 . upon deployment , the nodes broadcast their radius . all nodes reset their transmission radius to be the maximum of their original radius and the ones they receive from the broadcast . note that this is done only once . clearly all the directed links in the original graph now become bi - directional together with possible creation of some directed edges . thus if the enhanced graph @xmath45 is connected , then so is the graph obtained by this procedure . for any fixed @xmath46 , define the sequence of _ cut - off _ functions @xmath47 , via the equation @xmath48 later we will have occasion to take @xmath49 to be a function of @xmath4 and @xmath6 as well . we will denote @xmath50 by @xmath51 and the associated enhanced graph by @xmath45 , when @xmath49 , @xmath2 are fixed and @xmath52 is as defined in ( [ eqncutoff1 ] ) . by the palm theory for poisson point processes ( theorem 1.6 , @xcite ) , the expected out - degree of any node in @xmath51 will be @xmath53 = n \int_{b(x , r_n(c , x ) ) } f(y ) \,dy = c \log n,\vadjust{\goodbreak}\ ] ] which is the same as the vertex degree in the usual random geometric graph defined on uniform points in the connectivity regime . let @xmath54 and @xmath55 denote the palm distributions of @xmath29 conditional on a vertex located at @xmath4 . by the palm theory for poisson point processes , the expected out - degree of a node located at @xmath32 in the graph @xmath51 will be @xmath56 = c \log n.\ ] ] thus the expected vertex degree of a node in @xmath51 does not depend on the location of the node . in fact , the number of points of @xmath57 that fall in @xmath58 under @xmath54 will follow a poisson distribution with mean @xmath59 . let @xmath60 be the number of nodes in @xmath51 that have zero out - degree , that is , @xmath61 = 0\}}. \label{eqnwn}\ ] ] for each @xmath39 , define @xmath62 in other words , @xmath63 is the critical cut - off parameter , that is , the smallest @xmath49 , so that the graphs @xmath50 do not have any node with zero out - degree . our first result is a strong law for this critical cut - off parameter @xmath63 . [ thmmdnstar ] let @xmath63 be the critical cut - off parameter as defined in ( [ eqncriticald ] ) . then almost surely , @xmath64 let @xmath65 be the random geometric graphs as defined in definition [ defrgg ] with @xmath52 as in ( [ eqncutoff1 ] ) . consider the enhanced random geometric graph @xmath66 associated with @xmath51 ; see definition [ defrgge ] . we now state a strong law result for the critical cut - off parameter to eleminate isolated nodes in the graph @xmath45 . let @xmath67 be the number of isolated nodes , that is , nodes with degree zero , in the enhanced graph @xmath45 . define @xmath68 clearly @xmath69 by construction . the following theorem shows that the threshold required to eleminate isolated nodes in the enhanced graph @xmath45 is the same as for the graph @xmath51 . [ thmmtdn ] let @xmath70 be as defined in ( [ eqntdn ] ) . then , almost surely , @xmath71 the exact asymptotics for the connectivity threshold for random geometric graphs requires a lot of elaborate computations ; see chapter 13 , @xcite . we provide a sufficient condition that requires only a local computation at each node and makes use of the connectivity threshold for the usual uniform random geometric graphs . let @xmath17 be a random variable with probability density function @xmath2 with support @xmath72 @xmath15 . suppose that @xmath2 admits a mapping @xmath73 such that @xmath74 is uniformly distributed on @xmath10^d$ ] . for example , if the coordinates of @xmath17 are independently distributed , then the coordinate mappings of @xmath75 will be the marginal distributions . recall that @xmath22 denotes the volume of the unit ball . for any @xmath76 , let @xmath77 be the sequence defined by @xmath78 where @xmath79 for any set @xmath24 , we denote by @xmath80 the image of the set @xmath24 under @xmath75 . suppose that the functions @xmath81 are as defined in ( [ eqncutoff1 ] ) . define the sequence of functions , @xmath82 and @xmath83 to be functions on @xmath84 that satisfy the equation @xmath85 let @xmath86 be the graphs defined as in definition [ defrgg ] with @xmath34 replaced by @xmath87 . let @xmath88 be the enhanced graphs associated with @xmath89 , that is , the graphs obtained by making the edges in @xmath86 bi - directional . [ thmmconnectivity ] let @xmath18 , and suppose that @xmath74 is uniform on @xmath10^d$ ] , . then for any @xmath76 , almost surely , the sequence of enhanced random geometric graphs @xmath90 is connected for all sufficiently large @xmath6 . in particular , the above result implies that @xmath91 our next result is on strong law asymptotics for the maximum and minimum vertex degrees for the sequence of graphs @xmath51 . let @xmath92 be defined by @xmath93 and @xmath94 the function @xmath95 has a unique turning point at the minima @xmath96 . let @xmath97 be the inverse of @xmath95 restricted to @xmath98 and @xmath99\rightarrow[0,1]$ ] be the inverse of the restriction of @xmath95 to @xmath10 $ ] . [ thmmmaxmindeg ] for any @xmath46 , let @xmath100 be the maximum and @xmath101 be the minimum vertex out - degree of the graph @xmath102 then with probability 1 , @xmath103 if @xmath104 , then @xmath105 almost surely . if @xmath106 , then with probability 1 , @xmath107 note that ( [ eqndeln ] ) does not shed any light on what happens to the minimum vertex degree when @xmath108 . this requires a finer parametrization of the cut - off function . for any @xmath109 , define @xmath110 to be functions satisfying @xmath111 for all n sufficiently large for which @xmath112 , and arbitrarily otherwise . since @xmath113 is fixed , we will write @xmath114 for @xmath115 . for each @xmath39 , define the sets @xmath116 [ thmmpoicgs ] let @xmath117 be the number of nodes of out - degree zero in the graph @xmath118 where @xmath119 is as defined in ( [ eqnrpoi ] ) . suppose that @xmath2 satisfies the following two conditions for all @xmath6 sufficiently large : there exists a constant @xmath120 , such that for all @xmath6 sufficiently large @xmath121 @xmath122 then , @xmath123 as @xmath124 , where @xmath125 denotes a poisson random variable with mean @xmath126 as noted earlier , poisson approximation results for the number of isolated nodes are available only for a small class of distributions . condition ( [ eqnc2 ] ) can be replaced by the following sufficient condition which , as will be shown below , is easier to verify for some classes of densities . @xmath127 [ thmmpoi2 ] suppose that the hypothesis ( [ eqnc1 ] ) of theorem [ thmmpoicgs ] and ( [ eqnc3 ] ) are satisfied , then ( [ eqnpoicgs ] ) holds true . our final result is an illustration of the use of the above theorem to a class of densities with compact support that have at most polynomial rates of decay to zero . let @xmath2 be a continuous density with compact support @xmath16 , @xmath15 . let @xmath128 , @xmath129 be nonintersecting balls such that @xmath130 on the boundary of these balls . for each @xmath131 , there exists integers @xmath132 and @xmath133 , @xmath134 and constants @xmath135 such that , either @xmath136 or @xmath137 and @xmath138 elsewhere in @xmath84 . the density can vanish , for example , over water bodies which may contain islands ( which are being approximated by balls ) , and the density decays to zero polynomially near the boundary of these balls . in particular , @xmath84 itself could be taken to be a ball with the density decaying polynomially to zero at the boundary of @xmath84 in a radially symmetric fashion . we will denote by @xmath139 the set of all densities of the above form . the class @xmath139 contains many of the standard distributions such as the uniform , triangular , beta , etc . , truncated versions of standard distributions with unbounded support such as the gaussian , gamma , etc . in santi @xcite the density is assumed to be continuous and bounded away from 0 over @xmath10 ^ 2 $ ] , which is contained in @xmath139 . the class @xmath139 also contains the higher dimensional extensions of the polynomial densities considered in foh et al . @xcite and han and makowski @xcite . as remarked earlier , since our motivation was to allow for nonstandard densities , @xmath139 contains functions that are the modulus of analytic functions in the interior of @xmath84 . [ corpoi ] the conclusion of theorem [ thmmpoi2 ] holds for any density in class @xmath139 . proof of theorem [ thmmdnstar ] first we will show that @xmath140 fix @xmath76 and let @xmath141 . let @xmath142 , @xmath143 , where the constant @xmath144 will be chosen later . let @xmath145 be as defined in ( [ eqnwn ] ) . for each @xmath146 define the events @xmath147 recall that @xmath148 $ ] denotes the number of points of the point process @xmath29 that fall in the set @xmath24 . set @xmath149 we will show that @xmath150 < \infty . \label{eqnsumbk}\ ] ] it will then follow by the borel cantelli lemma that almost surely , only finitely many of the events @xmath151 ( and hence the @xmath152 ) happen . consequently , with probability 1 , @xmath153 , eventually . since @xmath76 is arbitrary , this will prove ( [ eqnlimsupdnstar ] ) . for each @xmath154 let @xmath155 be the event that there is a vertex @xmath156 that has out - degree zero in the graph @xmath157 that is , @xmath158 = 0 \bigr\}.\ ] ] since we have assumed the variables @xmath159 to be nondecreasing and the functions @xmath81 are nonincreasing in @xmath6 for each fixed @xmath49 and @xmath4 , we have @xmath160 consequently , @xmath161 by the palm theory for poisson point processes ( theorem 1.6 , @xcite ) , ( [ eqnf ] ) and ( [ eqncutoff1 ] ) , we have @xmath162 & \leq & e\biggl[\sum_{x \in\cp_{n_{k+1 } } } 1_{\{\cp_{n_k}[b(x , r_{n_{k+1}}(c))\setminus\{x\ } ] = 0 \ } } \biggr ] \nonumber\\ & = & n_{k+1 } \int_{\mr^d } e^{-n_k f(b(x , r_{n_{k+1}}(c ) ) ) } f(x ) \,dx \nonumber \\[-8pt ] \\[-8pt ] \nonumber & = & n_{k+1 } \exp\biggl ( - c \frac{n_k}{n_{k+1 } } \log n_{k+1 } \biggr ) \\ & = & ( k+1)^b \exp\biggl ( - ( 1+\ep ) \biggl(\frac{k}{k+1 } \biggr)^b \log(k+1)^b \biggr).\nonumber\end{aligned}\ ] ] choose @xmath163 so that @xmath164 , and pick @xmath144 such that @xmath165 for sufficiently large @xmath166 , we have @xmath167 using this in ( [ eqnhk ] ) , we get for all @xmath168 sufficiently large @xmath169 \leq\frac{1}{(k+1)^{((1-\ga)(1+\ep ) - 1)b}},\ ] ] which is summable in @xmath166 . ( [ eqnsumbk ] ) now follows from the above inequality and ( [ eqnbk ] ) . this proves ( [ eqnlimsupdnstar ] ) by the arguments following ( [ eqnsumbk ] ) . to complete the proof , we need to show that @xmath170 fix @xmath104 and pick @xmath171 such that @xmath172 . choose @xmath173 such that @xmath174 since @xmath2 is continuous , we can and do fix a @xmath175 satisfying @xmath176 , where @xmath177 let @xmath178 be such that @xmath179 choose @xmath180 such that @xmath181 and let @xmath182 define the sequence of functions @xmath183 by @xmath184 let @xmath185 be the maximum number such that there exists @xmath185 many disjoint balls of radius @xmath186 with centers in @xmath187 . then ( see lemma 2.1 , @xcite ) , we can find a constant @xmath188 such that for all @xmath6 sufficiently large , @xmath189 let @xmath190 be the deterministic set of points in @xmath187 such that the balls @xmath191 , @xmath192 , are disjoint . let @xmath193 be the event that there is exactly one point of @xmath29 in @xmath194 with no other point in @xmath195 , that is , @xmath196 = 1 , \cp _ n[b(x,\bar{r}_n(u))\setminus b(x,\bar{r}_n(\ep ) ) ] = 0 \}. \label{eqnen}\vadjust{\goodbreak}\ ] ] note that the events @xmath197 = 1\}$ ] and @xmath198 = 0 \}$ ] are independent . hence for any @xmath199 , we have @xmath200 & = & n f(b(x,\bar{r}_n(\ep ) ) ) e^{- n f(b(x,\bar{r}_n(\ep ) ) ) } e^{- n f(b(x,\bar{r}_n(u)))\setminus b(x,\bar{r}_n(\ep ) ) } \\ & = & n f(b(x,\bar{r}_n(\ep ) ) ) e^{- n f(b(x,\bar{r}_n(u)))}.\end{aligned}\ ] ] note that for all @xmath199 and all @xmath6 sufficiently large , @xmath201 . hence for all @xmath6 sufficiently large and @xmath199 , we have @xmath202 & \geq & n f_0 \th_d \bar{r}_n(\ep)^d \exp ( - n g_0 \th _ d \bar{r}_n(u)^d ) \nonumber\\ & = & \ep\log n \exp\biggl ( - \frac{g_0 u}{f_0 } \log n \biggr ) \\ & = & \ep n^{- { g_0u}/{f_0 } } \log n.\nonumber\end{aligned}\ ] ] using the fact that the events @xmath203 , @xmath204 , are independent , the inequality @xmath205 , and ( [ eqnsg ] ) , ( [ eqnlben ] ) , we get @xmath206 \leq\exp\bigl ( - \ep\sg_n n^{- { ( g_0u)}/{f_0 } } \log n \bigr ) \leq\exp\bigl ( - c_1 \ep n^{1 - { ( g_0u)}/{f_0 } } \bigr),\ ] ] which is summable in @xmath6 since @xmath207 hence , by the borel cantelli lemma , almost surely , for all sufficiently large @xmath6 the event @xmath203 happens for some @xmath208 . hence w.p . 1 , for all n sufficiently large , we can find a random sequence @xmath209 such that @xmath210 , and there is no other point of @xmath29 within a distance @xmath211 of @xmath212 . by ( [ eqnep ] ) , @xmath213 since @xmath214 , from ( [ eqncutoff1 ] ) and the remark above ( [ eqnlben ] ) , we get for sufficiently large @xmath6 , @xmath215 \\[-8pt ] \nonumber & = & \th_d f_0 \bar{r}_n(c)^d \leq \int_{b(x_{j(n)},\bar{r}_n(c ) ) } f(y ) \,dy.\end{aligned}\ ] ] hence @xmath216 . it follows that there is no other point of @xmath29 in @xmath217 , that is , @xmath212 has out - degree zero in @xmath218 consequently w.p . 1 , @xmath219 for all @xmath6 sufficiently large . since @xmath220 , this proves ( [ eqnliminfdnstar ] ) . proof of theorem [ thmmtdn ] since the graph @xmath45 is obtained by making all the edges in @xmath51 bi - directional , @xmath221 . consequently @xmath69 , and hence by theorem [ thmmdnstar ] we have @xmath222 thus it suffices to show that @xmath223 fix @xmath220 , and let @xmath224 and @xmath225 be as in the second part of the proof of theorem [ thmmdnstar ] . recall the random sequence @xmath209 , defined on a set of probability one , such that for sufficiently large @xmath6 , the point @xmath210 has no other point of @xmath29 within a distance @xmath226 . since @xmath181 , by the same arguments as in ( [ eqnrnbound ] ) , we have @xmath227 for all @xmath228 and all @xmath6 sufficiently large . thus almost surely , none of the points @xmath229 , that fall in the ball @xmath230 , have an out - going edge to @xmath212 for all sufficiently large @xmath6 . on the other hand , for any point @xmath231 , suppose @xmath232 . then we can find a point @xmath8 such that @xmath233 such that @xmath234 however , @xmath235 which clearly is not possible for sufficiently large @xmath6 . thus there can be no edge leading from a point of @xmath29 in @xmath236 to @xmath237 hence the points @xmath210 have zero in - degree as well for sufficiently large @xmath6 . consequently , with probability 1 , @xmath238 for all sufficiently large @xmath6 for any @xmath239 this proves ( [ eqnliminftdn ] ) and thus completes the proof of theorem [ thmmtdn ] . proof of theorem [ thmmconnectivity ] for any @xmath76 , let @xmath240 , @xmath39 , be as defined in ( [ eqncn ] ) . consider the random geometric graph @xmath241 induced by the mapping @xmath75 and the enhanced graph @xmath242 . the vertex set of the graph @xmath241 is the set @xmath243 with edges between any two vertices @xmath244 and @xmath245 provided there is an edge between @xmath246 and @xmath247 in the graph @xmath45 . the vertices of @xmath241 are distributed according to a homogenous poisson point process on @xmath10^d$ ] with intensity @xmath6 . let @xmath248 be as defined in ( [ eqnm ] ) . now by theorem 13.2 , @xcite ( with @xmath249 ) , we have @xmath250 almost surely , where @xmath251 is the threshold for simple connectivity in the usual uniform random geometric graph on @xmath10^d$ ] ( nodes being distributed according to a homogenous poisson point process with intensity @xmath6 ) . by definition of @xmath252 , in the graph @xmath241 , each vertex is connected to all its neighbors that are within a distance @xmath253 almost surely for all sufficiently large @xmath6 . it follows that almost surely , the graphs @xmath241 , and hence the graphs @xmath45 are connected for all sufficiently large @xmath6 . proof of theorem [ thmmmaxmindeg ] we first show ( [ eqndeln ] ) . fix @xmath46 and @xmath254 define the sequence @xmath255 fix a constant @xmath144 such that @xmath256 , and let @xmath257 @xmath258 define the events @xmath259 , @xmath39 . let @xmath260 since @xmath159 , @xmath261 are increasing , and @xmath81 is decreasing pointwise in @xmath6 , we have @xmath262 \geq c_{n_k } + 1 \}. \label{eqnmaxbk}\ ] ] by the palm theory for poisson point processes ( theorem 1.6 , @xcite ) , @xmath263 \geq c_{n_k } \ } } \biggr ] \nonumber \\[-8pt ] \\[-8pt ] \nonumber & = & n_{k+1 } \int_{\mr^d } p \bigl ( \operatorname{po } ( n_{k+1 } f(b(x , r_{n_k}(c , x ) ) ) ) \geq c_{n_k } \bigr ) f(x ) \,dx,\end{aligned}\ ] ] where @xmath125 denotes a poisson random variable with mean @xmath126 from ( [ eqncutoff1 ] ) , we have for sufficiently large @xmath166 , @xmath264 hence for sufficiently large @xmath166 , by definition of @xmath265 , we get @xmath266 \\[-8pt ] \nonumber & = & h_+^{-1}\biggl ( \frac{1+\ep}{c } \biggr ) > 1.\end{aligned}\ ] ] hence using the chernoff bound for the poisson distribution ( see lemma 1.2 , @xcite ) , we get @xmath267 since @xmath95 is increasing in @xmath268 and @xmath269 , using ( [ eqnratio ] ) we can bound the probability on the left - hand side in the above equation by @xmath270 substituting this bound in ( [ eqnmaxbbd ] ) , we get for sufficiently large @xmath166 , @xmath271 which is summable in @xmath166 , since @xmath272 hence , by the borel cantelli lemma , almost surely only finitely many of the events @xmath151 and hence @xmath152 happen . hence , almost surely , @xmath273 eventually . the result now follows since @xmath76 is arbitrary . the result for @xmath274 in case @xmath104 follows from theorem [ thmmdnstar ] . for the case @xmath275 the proof of ( [ eqndeln ] ) is entirely analogous to that of ( [ eqndeln ] ) , and so we provide the corresponding expressions . fix @xmath276 such that @xmath277 . let @xmath278 be such that @xmath256 , and define the sequence @xmath142 , @xmath143 . define the events @xmath279 where @xmath280 let the events @xmath151 be as defined in ( [ eqndefbk ] ) . the expression analogous to ( [ eqnmaxbk ] ) will be @xmath281 \leq c_{n_{k+1 } } + 1 \}. \label{eqnminbk}\ ] ] for sufficiently large @xmath166 , @xmath282 hence for sufficiently large @xmath166 , by our choice of @xmath283 , we get @xmath284 \\[-12pt ] \nonumber & = & h_-^{-1 } \biggl ( \frac{1+\ep}{(1 - \ep)c } \biggr ) < 1.\hspace{-35pt}\end{aligned}\ ] ] again using the chernoff bound and proceeding as in the previous proof , we will get @xmath285 which is summable in @xmath166 . since @xmath76 is arbitrary , ( [ eqndeln ] ) now follows by the borel cantelli lemma and the arguments used earlier to infer ( [ eqndeln ] ) . the proof of theorem [ thmmpoicgs ] uses the following lemma , which is a straightforward extension of theorem 6.7 , @xcite . let @xmath286 denote the total variation distance between two random variables . let @xmath287 and @xmath288 be as in theorem [ thmmpoicgs ] . we will use the notation , @xmath289 , @xmath290 and @xmath291 recall that @xmath292 denotes the number of points of @xmath29 that fall in the set @xmath24 . [ lempoiapprox ] let @xmath2 be a continuous density on @xmath3 . then , for the graph @xmath293 , we have @xmath294 ) ) \leq \min\biggl ( 3 , \frac{1}{e[\hw_n ] } \biggr ) \bigl(i_n^{(1 ) } + i_n^{(2)}\bigr ) , \label{eqnpoiapprox}\ ] ] where @xmath295 \\[-8pt ] \nonumber & & { } \times p [ \cp_n(b_n(x ) ) = 0 ] p [ \cp_n(b_n(y ) ) = 0 ] , \\ % \label{eqni2}i_n^{(2 ) } & = & n^2 \int_{\mr^d } f(x ) \,dx \int_{\hat{b}_n(x ) } f(y ) \,dy \nonumber \\[-8pt ] \\[-8pt ] \nonumber & & { } \times p [ \cp_n^y(b_n(x ) ) = 0 , \cp_n^x(b_n(y ) ) = 0].\end{aligned}\ ] ] the proof is identical to the proof of theorem 6.7 , @xcite with the following obvious change . in the definition of dependency neighborhood , the parameter @xmath34 is replaced by the function @xmath296 . proof of theorem [ thmmpoicgs ] by the palm theory for poisson point processes , @xmath297 = n \int_{\mr^d } f(x ) \,dx\ , e^{- n \int_{b_n(x ) } f(y ) \,dy } , \ ] ] where @xmath298 is as defined above lemma [ lempoiapprox ] . using ( [ eqnrpoi ] ) , we get @xmath297 = e^{-\be}.\ ] ] hence , by lemma [ lempoiapprox ] , it suffices to show that @xmath299 converge to zero as . again , using the palm theory and ( [ eqnrpoi ] ) , we get @xmath300 \\[-8pt ] \nonumber & = & e^{-2 \be } \int_{\mr^d } f(x ) \,dx \int_{\bb_n(x ) } f(y ) \,dy \to 0\hspace{-35pt}\end{aligned}\ ] ] as @xmath124 , by the dominated convergence theorem , since @xmath301 for each @xmath32 . next we will show that @xmath302 , as @xmath124 . @xmath303 \\[-8pt ] \nonumber & = & i_n^{(21 ) } + i_n^{(22)},\end{aligned}\ ] ] where @xmath304 where @xmath305 is as defined in ( [ eqnhan ] ) . consider the inner integral in @xmath306 . on the set @xmath307 , we have @xmath308 , and hence by ( [ eqnrpoi ] ) @xmath309 thus , @xmath306 converges to zero using the same arguments as in ( [ eqnin1 ] ) . it remains to show that @xmath310 as @xmath124 . since @xmath311 , where @xmath312 , we get from ( [ eqnrpoi ] ) , ( [ eqnc1 ] ) , that @xmath313 is bounded by a constant times @xmath314 where @xmath315 is as defined in ( [ eqnan ] ) . the expression in ( [ eqnfinalbound ] ) converges to zero as @xmath124 by ( [ eqnc2 ] ) . proof of theorem [ thmmpoi2 ] note that in the proof of theorem [ thmmpoicgs ] , ( [ eqnc2 ] ) is used only to prove that ( [ eqnfinalbound ] ) converges to zero . hence it suffices to show that ( [ eqnfinalbound ] ) converges to zero under ( [ eqnc3 ] ) . let @xmath315 be as defined in ( [ eqnan ] ) . then @xmath316 using this , the expression in ( [ eqnfinalbound ] ) is bounded by @xmath317 applying fubini s theorem to the second term above , we see that the above expression is bounded by @xmath318 which converges to zero by ( [ eqnc3 ] ) . proof of corollary [ corpoi ] let @xmath319 . consider the following two classes of densities : @xmath320 where @xmath321 is the support of @xmath2 . denote by @xmath322 the set of functions @xmath323 such that for some @xmath324 and @xmath325 , @xmath326 and @xmath327 @xmath328 with @xmath329 . let @xmath330 be the set of functions @xmath323 such that for some @xmath325 , @xmath331 @xmath332 with @xmath329 . we first prove the result for densities @xmath333 . to do this we need to verify conditions ( [ eqnc1 ] ) and ( [ eqnc3 ] ) . _ step _ 1 . fix @xmath334 . let @xmath335 and @xmath336 and @xmath337 . by ( [ eqnrpoi ] ) , for any @xmath338 and for all @xmath6 large enough , we get @xmath339 or @xmath340 to prove ( [ eqnc1 ] ) , note that for any @xmath341 , we can inscribe a ball of radius @xmath342 inside @xmath343 . from this and ( [ boundsrn ] ) , for all @xmath6 sufficiently large , we get @xmath344 this proves ( [ eqnc1 ] ) with @xmath345 . by ( [ boundsrn ] ) we have @xmath346 this proves ( [ eqnc3 ] ) . thus the poisson convergence result holds for any @xmath334 . _ step _ 2 . next we prove the result for @xmath347 . proof for @xmath348 is similar and so we omit it . let @xmath349 . recall that @xmath350 over @xmath351 and is of the form @xmath352 over @xmath353 . for any @xmath354 , we have by ( [ eqnrpoi ] ) , @xmath355 note that the density is radially increasing , that is , @xmath356 if @xmath357 . if @xmath341 , then @xmath358 . if @xmath357 , then using ( [ eqnequalityoverball ] ) and the monotonicity of @xmath2 we get @xmath359 on the other hand if @xmath341 and @xmath360 , then by ( [ eqnequalityoverball ] ) and the monotonicity of @xmath2 we have @xmath361 hence @xmath362 thus ( [ eqnc1 ] ) holds with @xmath363 . next we verify ( [ eqnc3 ] ) over @xmath353 . by ( [ eqnrpoi ] ) , we have @xmath364 by changing to polar coordinates , the integral on the right - hand side of the above equation has the bounds @xmath365 for some positive constants @xmath366 from ( [ eqnfr ] ) and ( [ eqnboundr ] ) , we get @xmath367 and hence for some constant @xmath49 , @xmath368 this proves ( [ eqnc3 ] ) . the result now follows for any @xmath369 by theorem [ thmmpoi2 ] .	we propose a _ distribution - free _ approach to the study of random geometric graphs . the distribution of vertices follows a poisson point process with intensity function @xmath0 , where @xmath1 , and @xmath2 is a probability density function on @xmath3 . a vertex located at @xmath4 connects via directed edges to other vertices that are within a _ cut - off _ distance @xmath5 . we prove strong law results for ( i ) the critical cut - off function so that almost surely , the graph does not contain any node with out - degree zero for sufficiently large @xmath6 and ( ii ) the maximum and minimum vertex degrees . we also provide a characterization of the cut - off function for which the number of nodes with out - degree zero converges in distribution to a poisson random variable . we illustrate this result for a class of densities with compact support that have at most polynomial rates of decay to zero . finally , we state a sufficient condition for an enhanced version of the above graph to be almost surely connected eventually . .
among the alternative approaches to describe the dark cosmological sector , consisting of dark matter ( dm ) and dark energy ( de ) , so called holographic de models have received considerable attention @xcite . the underlying holographic principle states that the number of degrees of freedom in a bounded system should be finite and related to the area of its boundary @xcite . on this basis , a field theoretical relation between a short distance ( ultraviolet ) cutoff and a long distance ( infrared ) cutoff was established @xcite . such relation ensures that the energy in a box of size @xmath3 does not exceed the energy of a black hole of the same size . if applied to the dynamics of the universe , @xmath3 has to be a cosmological length scale . different choices of this cutoff scale result in different de models . for the most obvious choice , the hubble scale , only models in which dm and de are interacting with each other also nongravitationally , give rise to a suitable dynamics @xcite . following ref . , there has been a considerable number of investigations based on the future event horizon as cutoff scale . all models with a cutoff at the future event horizon , however , suffer from the serious drawback that they can not describe a transition from decelerated to accelerated expansion . a future event horizon does not exist during the period of decelerated expansion . we are focussing here on a further option that has received attention more recently , a model based on a cutoff length proportional to the ricci scale . a distance proportional to the ricci scale has been identified as a causal connection scale for perturbations @xcite . as a cutoff length in de models it was first used in ref . . subsequent investigations include ref . and , for the perturbation dynamics , refs . + here , we reconsider the dynamics of a two - component system of pressureless dm and ricci - type de both in the homogeneous and isotropic background and on the perturbative level@xcite . whereas most dynamic de scenarios start with an assumption for the equation - of - state ( eos ) parameter for the de , the starting point of holographic models is an expression for the de energy density from which the eos is then derived . as was pointed out in ref . , the mere definition of the holographic de density , independently of the choice of the specific cutoff length , implies an interaction with the dm component . requiring this interaction to vanish is equivalent to impose an additional condition on the dynamics . in the case of ricci - type de this condition establishes a simple relation between the matter fraction and the necessarily time - dependent eos parameter . of course , a time - varying eos parameter is not compatible with a cosmological constant . our main aim here is to perform a gauge - invariant perturbation analysis for this model . it will turn out that the general perturbation dynamics suffers from instabilities . there exists just one single configuration without instabilities at finite values of the scale factor @xmath4 @xcite . we also update previous tests of the homogeneous and isotropic background dynamics using recent results for the differential age of old objects based on the @xmath5 dependence , data from snia and from bao . the cosmic medium is assumed to consist of pressureless dm with energy density @xmath6 and a holographic de component with energy density @xmath7 . in the spatially flat case friedmann s equation is @xmath8 where @xmath9 is the hubble rate and @xmath4 is the scale factor of the robertson - walker metric . in general , both components are not separately conserved but obey the balance equations @xmath10 with a source ( or loss ) term @xmath11 , such that the total energy @xmath12 is conserved . here , @xmath13 is the eos parameter of the de and @xmath14 is the pressure associated with the holographic component . in terms of the scale factor as independent variable , the acceleration equation can be written @xmath15 where @xmath16 is the ratio of the energy densities . total _ effective eos of the cosmic medium is @xmath17 according to the balance equations ( [ cons2 ] ) , the ratio @xmath18 changes as @xmath19\ . \label{dr2}\ ] ] following refs . , we write the holographic energy density as @xmath20 the quantity @xmath3 is the infrared cutoff scale and @xmath21 is the reduced planck mass . the numerical constant @xmath22 determines the degree of saturation of the condition @xmath23 which is at the heart of any holographic de model . it states that the energy in a box of size @xmath3 should not exceed the energy of a black hole of the same size @xcite . differentiation of the expression ( [ ans ] ) for the holographic de density and use of the energy balances ( [ cons2 ] ) yields @xmath24 in general , there is no reason for @xmath11 to vanish . assuming @xmath25 provides us with a specific relationship between @xmath26 and the ratio of the rates @xmath27 and @xmath28 . any nonvanishing @xmath11 will modify this relationship . with @xmath11 from ( [ ql ] ) , the general dynamics ( [ dr2 ] ) of the energy - density ratio @xmath18 becomes @xmath29\ , . \label{drl}\ ] ] the case without interaction is characterized by [ cf . ( [ dr2 ] ) ] @xmath30 with a generally time - dependent @xmath26 . different choices of the cutoff scale @xmath3 give rise to different expressions for the total effective eos parameter in eq . ( [ w ] ) and to different relations between @xmath26 and @xmath18 . we shall briefly sketch the situations for the hubble radius and for the future event horizon as cutoff lengths before considering in detail the ricci scale . for @xmath31 the holographic de density is @xmath32 for the deceleration parameter one has @xmath33 in the interaction - free case we recover the einstein - de sitter value @xmath34 . the condition for accelerated expansion is @xmath35 . to describe a transition from decelerated to accelerated expansion , @xmath11 has to change from @xmath36 to @xmath35 . a viable scenario can be realized , e.g. , by a choice @xcite @xmath37 where @xmath38 is an interaction constant . the resulting dynamics is that of a generalized chaplygin gas with a hubble rate @xmath39^{1/n}\ , , \label{hq}\ ] ] where @xmath38 is related to the present value @xmath40 of the deceleration parameter @xmath41 by @xmath42 for @xmath43 one reproduces the @xmath2cdm dynamics . with @xmath44 , where @xmath45 is the future event horizon , the holographic de density ( [ ans ] ) is @xmath46 the de balance ( [ cons2 ] ) can be written as @xmath47 with an effective eos ( the superscript e denotes the event horizon ) @xmath48 this effective eos does not directly depend on @xmath26 . however , the ratio @xmath18 that enters @xmath49 is determined by @xmath26 via eq . ( [ drl ] ) . notice also , that this effective eos for the de component is different from the total effective eos of the cosmic medium which is @xmath50 . in the previous hubble - scale - cutoff case both these quantities were identical . different from the previous hubble - scale cutoff , there exists a non - interacting limit with accelerated expansion in the present case . in this special situation @xmath51\ \mathrm{and } \ r_{e}h = - \frac{2}{1 + 3 \omega } \label{q0}\ ] ] are valid . explicitly , the relation between @xmath26 and @xmath18 is @xmath52 it is obvious , that for any @xmath53 , the parameter @xmath26 remains always smaller than @xmath54 , demonstrating the impossibility of a matter - dominated period in this context . our interest in the present paper will be the ricci - scale cutoff . the role of a distance proportional to the ricci scale as a causal connection scale for perturbations was noticed in ref . . in ref . it was used for the first time as a de cutoff scale . the ricci scalar is @xmath55 . for the corresponding cutoff scale one has @xmath56 , i.e. , @xmath57 where @xmath58 . upon using ( [ dh ] ) we obtain @xmath59 for the holographic de density . notice that the ( not yet known ) eos parameter explicitly enters @xmath60 . use of friedmann s equation provides us with @xmath61 which coincides with the result in ref . . obviously , a constant value of @xmath26 necessarily implies a constant @xmath18 and vice versa . for the source term we have @xcite @xmath62\rho_{h } \ , . \label{q=}\ ] ] it is obvious that a constant eos parameter @xmath26 is compatible with @xmath25 only for @xmath63 , i.e. , if @xmath7 behaves as dust . for a time - varying eos parameter @xmath64 , however , there exists a non trivial case @xmath25 : @xmath65 in the remainder of the paper we shall consider this case , for which the eos parameter is explicitly given by @xmath66 at high redshift we have @xmath67 the property that noninteracting ricci - de behaves as dust at high redshift was first pointed out in ref . . this model naturally reproduces an early matter - dominated era . for @xmath68 and @xmath69 , the ratio @xmath18 approaches @xmath70 for @xmath71 . this value is only roughly ten times larger than the present value @xmath72 . for the @xmath2cdm model the corresponding difference is about nine orders of magnitude . in this sense , the coincidence problem is considerably alleviated for our ricci cdm model for which the hubble rate becomes @xmath73 } { 3\omega_{0}- r_{0}\left[1+r_{0}-3\omega_{0}\right ] } } \ , . \label{solhq0}\ ] ] generally , the two - component model is described by an energy - momentum tensor @xmath74 with @xmath75 and @xmath76 . the quantity @xmath77 denotes the total four - velocity of the cosmic substratum . latin indices run from @xmath78 to @xmath79 . the total @xmath80 splits into a matter component and a holographic de component , @xmath81 with ( @xmath82 ) @xmath83 for separately conserved fluids we have @xmath84 and @xmath85 . in general , each component has its own four - velocity , with @xmath86 . for the homogeneous and isotropic background we assume @xmath87 . indicating first - order perturbations about the homogeneous and isotropic background by a hat symbol , the perturbed time components of the four - velocities are @xmath88 restricting ourselves to scalar perturbations , we define the ( three- ) scalar quantities @xmath89 , @xmath90 and @xmath91 by @xmath92 introducing fractional energy - density perturbations @xmath93 and changing to gauge - invariant variables according to @xmath94 energy and momentum conservations for the cosmic medium as a whole reduce to @xcite @xmath95 the superscript c indicates that the corresponding variables are defined with respect to a comoving observer . the perturbation @xmath96 has to be determined from the perturbed raychaudhuri equation for @xmath97 . in our context one finds at linear order @xmath98 with @xmath99 where @xmath100 is the three - dimensional laplacian . combing eqs . ( [ balcomb ] ) and ( [ dthetacfin ] ) and specifying the pressure perturbations will result in a second - order equation for @xmath101 . on the other hand , with @xmath102 and @xmath103 it will be useful to consider the combination @xmath104 , where @xmath105 . our aim is to obtain an equation also for @xmath106 . to this purpose we have to introduce a further ingredient . so far the pressure perturbations are not sufficiently specified . in general , pressure perturbations in two - component systems are nonadiabatic . firstly , because of the two - component nature itself , secondly because each of the components may be nonadiabatic on its own . the relevant combination for our de component is @xmath107\,,\ ] ] which is a gauge - invariant expression . now an assumption for the perturbed eos parameter @xmath108 is necessary to proceed . we shall restrict ourselves here to adiabatic internal perturbations of the de component , equivalent to a vanishing of the combination ( [ hatph-1 ] ) : @xmath109 this assumption of an adiabatic de component allows us to relate the otherwise undetermined perturbation @xmath108 of the eos parameter to the de energy perturbation @xmath110 . we emphasize that the total perturbation dynamics remains nonadiabatic due to the two - component nature of the medium . the resulting coupled equations for @xmath101 and @xmath106 in the @xmath111 space then are @xcite ( the prime denotes a derivative with respect to @xmath4 ) @xmath112 \frac{\delta^{c\prime}}{a}\nonumber\\ & - & \left[\frac{3}{2 } + 12\frac{p}{\rho } - \frac{9}{2}\frac{p^{2}}{\rho^{2 } } - 9\frac{p^{\prime}}{\rho^{\prime } } - \frac{k^{2}}{a^{2}h^{2}}\frac{p}{\rho^{\prime}}\right]\frac{\delta^{c}}{a^{2}}\nonumber\\ & & \qquad\qquad\qquad\qquad = \frac{k^{2}}{a^{2}h^{2}}\frac{p^{\prime}}{\rho^{\prime}}\frac{\rho_{m}}{\rho}\frac{s_{mh}}{a^{2 } } \ \label{prprdeltas}\end{aligned}\ ] ] and @xmath113\frac{s^{\prime}_{mh}}{a}\nonumber\\ & + & \frac{r}{1+\omega}\frac{p}{\rho}\frac{k^{2}}{a^{2}h^{2}}\frac{s_{mh}}{a^{2 } } = \frac{1+r}{1+\omega}\frac{p}{\rho}\frac{k^{2}}{a^{2}h^{2}}\frac{\delta^{c}}{a^{2}}\,,\end{aligned}\ ] ] respectively , where we have to exclude the case @xmath114 . to obtain the matter - energy perturbations , we decompose the total energy - density perturbation @xmath101 according to @xmath115 combination with @xmath116 leads to @xmath117\,,\ ] ] which describes the matter - energy perturbations as a combination of @xmath101 and @xmath106 . to obtain its dynamics one has to solve the coupled system of equations ( [ prprdeltas ] ) and ( [ prprs ] ) . the matter density perturbation @xmath118 in relation ( [ deltamc ] ) is defined with respect to the _ total _ comoving gauge . to obtain the matter density perturbation , comoving with the matter velocity , @xmath119 , we have to consider @xmath120 since @xcite @xmath121 the quantity of interest is @xmath122 obviously , @xmath118 and @xmath123 differ by the last term in relation ( [ deltamcm ] ) . because of the factor @xmath124 ( assuming @xmath125 ) one expects that on scales smaller than the hubble scale the differences between @xmath118 and @xmath123 are small . ( dotted line ) , @xmath101 ( dash - dotted line ) , @xmath126 ( dashed line ) and @xmath127 ( thin solid line ) for @xmath128 , @xmath129 and @xmath130 . for comparison the @xmath2cdm result ( thick solid line ) is also included . ] in fig . [ fig1 ] we show the behavior of the quantities @xmath106 , @xmath101 , @xmath118 and @xmath131 for @xmath132 with @xmath133 . while this figure confirms that differences between @xmath118 and @xmath123 are indeed small on the chosen scale , there appear oscillations of all the perturbation quantities very close to the present time . this reminds of a similar feature in ( generalized ) chaplygin - gas models which apparently have jeopardized these models @xcite . still more serious is the existence of instabilities at future values @xmath134 of the scale factor , related to a crossing of the phantom divide @xmath114 . instabilities occur if the denominator @xmath135 in ( [ prprs ] ) vanishes , i.e. , if @xmath26 approaches @xmath136 . from eq . ( [ wsol ] ) one finds the condition for @xmath137 , @xmath138 which determines the value @xmath139 of the scale factor at which the instability occurs . solving for @xmath139 yields @xmath140 now we assume @xmath141 and consider the cases @xmath142 and @xmath143 separately . for @xmath144 and @xmath145 we have @xmath146 for @xmath147 we find @xmath148 , i.e. , the instability sets in at a finite value of the scale factor in the future . for @xmath149 , i.e. for a phantom eos , there appears an instability in the past at @xmath150 . since such kind of instability has not been observed , a present phantom eos is definitely excluded in the context of our model . the limit between the two regimes is just @xmath151 where we have @xmath152 , i.e. , an instability at the present epoch . the only case without instabilities at finite values of the scale factor is a fixed relation @xmath153 between the initially independent values of @xmath72 and @xmath1 . since @xmath154 necessarily , this implies @xmath155 . consequently , the only physically acceptable case is @xmath156 the parameters @xmath1 and @xmath72 are necessarily related to each other and can not be chosen independently . in a sense , @xmath72 quantifies the deviation of @xmath1 from @xmath157 . under this condition we have @xmath158 . this is exactly the result found by karwan and thitapura in their study of instabilities through nonadiabatic perturbations in a system of matter and ricci de @xcite . the solutions for @xmath26 and @xmath18 then simplify to @xmath159 respectively . combination of both solutions has the important consequence @xmath160 this makes all the coupling terms ( and some others ) in the coupled system ( [ prprdeltas ] ) and ( [ prprs ] ) vanish . also the pressure perturbations @xmath161 vanish . the square of the hubble rate turns out to be @xmath162 notice that we have the same number of free parameters as in the @xmath2cdm model , but there is no @xmath2cdm limit of ( [ hstab ] ) . the behavior of the perturbation quantities on the basis of ( [ wsol+ ] ) and ( [ dotp0 ] ) is visualized in fig . [ fig2 ] . this figure confirms that for the chosen configuration there are neither oscillations nor instabilities . from this point of view the model appears acceptable . ( dotted line ) , @xmath101 ( dash - dotted line ) , @xmath126 ( dashed line ) and @xmath127 ( thin solid line ) for @xmath129 and @xmath130 on the basis of ( [ wsol+ ] ) and ( [ dotp0 ] ) . the @xmath2cdm result is represented by the thick solid line . the relative density perturbations @xmath106 are negligible during the entire evolution . the results for @xmath126 and @xmath127 are almost identical . ] we have tested this model by the differential age of old objects based on the @xmath5 dependence @xcite as well as by the data from snia @xcite and from bao @xcite . the details of the analysis are given in ref . . the results are presented in fig . [ fig3 ] which shows the two - dimensional probability - distribution function ( pdf ) at @xmath163 ( @xmath164 of confidence level ) , @xmath165 ( @xmath166 of confidence level ) and @xmath167 ( @xmath168 of confidence level ) . the estimation for @xmath169 , based on a combination of the three tests at @xmath165 , is @xmath170 , while for @xmath171 we find @xmath172 . the straight line represents the combination @xmath173 which is singled out by the stability analysis of the perturbation dynamics . the tension to the results for the background dynamics is obvious , an agreement is possible only at the @xmath167 level . and @xmath171 resulting from a combination of the three tests . the straight line represents the instability - avoiding configuration @xmath173 . ] noninteracting ricci - type de is characterized by a necessarily time - dependent eos parameter . this makes it an observationally testable alternative to the @xmath2cdm model . there exists a a relationship between this eos parameter and the matter content of the universe . the ratio of the energy densities of dm and de varies considerably less than for the @xmath2cdm model . since the time of radiation decoupling it has changed by about one order of magnitude compared with roughly nine orders of magnitude for the @xmath2cdm model this amounts to a remarkable alleviation of the coincidence problem . ricci - type de behaves almost as dust at high redshift . our statistical analysis , based on recent observational data from snia , bao and @xmath5 , results in a preferred value of @xmath174 for the ricci - de parameter which confirms earlier studies in the literature @xcite . within a gauge - invariant analysis we calculated the matter perturbations as a combination of the total energy perturbations of the cosmic medium and the relative perturbations of the components . the perturbation dynamics suffers from instabilities that exclude a present phantom - type eos . it is only for a specific relation between the values @xmath0 of the present matter density and the present eos parameter @xmath1 that the dynamics remains stable for any finite scale - factor value . this relation corresponds to a ricci - de parameter @xmath175 @xcite . holographic ricci - type de represents a theoretically appealing scenario which does not need additional parameters except @xmath176 and @xmath0 . despite of its attractive features , the stable configuration is only marginally consistent with the observationally preferred background values of @xmath0 and @xmath1 . this work was supported by the comisin nacional de ciencias y tecnologa " ( chile ) through the fondecyt grants no . 1110230 and no . 1130628 ( r.h . and j.c.f and w.z acknowledge support by fondecyt - concurso incentivo a la cooperacin internacional " no . 1130628 as well as by cnpq ( brazil ) and fapes ( brazil ) . we thank alonso romero for carefully checking our calculations .	we consider the dynamics of a cosmological substratum of pressureless matter and holographic dark energy with a cutoff length proportional to the ricci scale . stability requirements for the matter perturbations are shown to single out a model with a fixed relation between the present matter fraction @xmath0 and the present value @xmath1 of the equation - of - state parameter of the dark energy . this model has the same number of free parameters as the @xmath2cdm model but it has no @xmath2cdm limit . we discuss the consistency between background observations and the mentioned stability - guaranteeing parameter combination .
with the running and upcoming experiments at lhc , rhic and gsi the analysis of the qcd phase diagram has become an important focus of research . not only the transition curves which separate different phases are of interest , but one would also like to understand from first principles the physical mechanisms that drive the various transitions . in this project we probe the finite temperature transition of qcd using static quark sources . in the framework of lattice qcd these can be implemented using local polyakov loops . the local polyakov loop @xmath1 is given by the trace of the product of temporal gauge links @xmath2 at a fixed spatial position @xmath3 ( @xmath4 is the number of lattice points in time direction ) : @xmath5 i.e. , the polyakov loop is a gauge transporter that propagates a static quark at position @xmath3 forward in time . for later use we also define the spatially averaged polyakov loop @xmath6 as @xmath7 , where @xmath8 is the spatial volume . due to translation invariance @xmath6 and @xmath1 have the same vacuum expectation value . after suitable renormalization the polyakov loop may be related to the free energy @xmath9 of a single quark , i.e. , @xmath10 , where @xmath11 is the temperature . below @xmath12 the free energy is infinite , thus @xmath13 , and the quarks are confined . above @xmath12 we have a finite free energy and thus @xmath14 , signaling deconfinement . hence the polyakov loop acts as an order parameter for deconfinement . in the deconfined phase of pure su(3 ) gauge theory the phases of the summed polyakov loops @xmath6 assume values in the vicinity of the three center elements @xmath15 of su(3 ) . this behavior shows the underlying center symmetry , a symmetry of the action and the path integral measure , which in pure gauge theory becomes broken spontaneously above @xmath12 . the polyakov loop @xmath6 transforms non - trivially under the center transformation and thus is also an order parameter for the symmetry breaking . above @xmath12 the polyakov loop @xmath16 is non - vanishing with phases near @xmath15 . for full qcd the fermion determinant breaks the center symmetry explicitly and acts as an external magnetic field favoring the real sector . except for the interchange of low and high temperature the situation is equivalent to simple spin systems with spins @xmath17 . without external field the corresponding hamiltonian has a symmetry which may be broken spontaneously . the symmetry breaking can be analyzed with observables that transform non - trivially under the symmetry group , e.g. , the magnetization @xmath18 . to obtain the equivalence between the spin system and the gauge theory one has to identify the local loops @xmath1 with the spins @xmath17 and the spatially averaged polyakov loop @xmath6 with the magnetization @xmath19 . in the absence of an external field ( the fermion determinant ) the phase of the magnetization @xmath20 ( the spatially averaged polyakov loop @xmath16 ) spontaneously selects one of the phases according to the underlying symmetry . if the external field ( the fermion determinant ) is turned on , the previously sharp transition becomes a crossover , and the phase of the magnetization ( the polyakov loop ) is determined by the symmetry breaking term . for the case of pure gauge theory these arguments are the basis of the svetitsky - yaffe conjecture @xcite which states that at @xmath12 pure su(3 ) gauge theory in 4 space - time dimensions can be described by a 3-d effective spin system with an effective action which is symmetric under the center group @xmath21 . the spin degrees of freedom are related to the local loops @xmath1 . the leading term of the effective action is given by ( @xmath22 and @xmath23 are real non - negative couplings ) @xmath24 \ , = \ , - \tau \sum_{\langle x , y \rangle } \big [ s(\vec{x})s(\vec{y})^ * \ , + \ , s(\vec{y})s(\vec{x})^ * \big ] \ ; - \ ; \kappa \sum_{\vec{x } } \big [ s(\vec{x } ) + s(\vec{x})^ * \big ] \ ; , \label{effact}\ ] ] where for illustration purposes we also included a symmetry breaking term which can be turned off when @xmath25 . in the simplest version the effective spins @xmath17 have values @xmath26 . magnetic finite temperature transitions for spin systems are well understood phenomena . for discrete symmetry groups the transition is accompanied by the formation of locally spin coherent weiss domains near @xmath12 , before one spin orientation wins out in the symmetry broken phase . even if a modest external magnetic field is applied one can still observe local clusters of aligned spins different from the direction preferred by the magnetic field . one can go a step further and analyze the connectedness properties of the weiss domains . one may define clusters of aligned spins and switch to a cluster description of magnetic systems @xcite . for several spin systems one finds that the magnetic transition may be characterized as a percolation phenomenon of suitably defined clusters . with the svetitsky - yaffe conjecture in mind , which describes the deconfinement transition with an effective spin system , we can formulate the central questions we explore in our project : * can one identify ( at least near @xmath12 ) characteristic properties of spin - like behavior in an ab - initio lattice simulation of pure gauge theory and/or full qcd ? * is it possible to identify spatial structures ( clusters ) that correspond to weiss domains ? * how do the domains behave near @xmath12 ? do suitably defined clusters percolate ? * what is the role of the fermion determinant which breaks the underlying center symmetry ? * can the clusters ( weiss domains ) be given a physical meaning also in the continuum limit ? for su(2 ) gauge theory similar questions were addressed in @xcite @xcite ( see also @xcite ) . first results for su(3 ) gauge theory were presented in @xcite . we study the polyakov loop ( [ ploopl ] ) using quenched configurations as well as configurations from full qcd . for our quenched analysis we use the lscher - weisz gauge action with lattice sizes from @xmath27 to @xmath28 and temperatures ranging from @xmath29 to @xmath30 @xcite . in full qcd we use configurations generated with a symanzik improved gauge action and @xmath31 flavors of stout - link improved staggered quarks at physical masses @xcite . we study lattices with sizes @xmath32 and @xmath33 in a temperature range from @xmath34 mev to @xmath35 mev . to analyze the properties of local polyakov loops we evaluate the @xmath1 and write them as @xmath36 i.e. , we decompose the local loops into modulus and phase . we begin with studying histograms @xmath37 $ ] and @xmath38 $ ] for the distribution of the local modulus @xmath39 and the local phase @xmath40 , shown in figs . 1 and 2 . in both figures the lhs . is for the quenched case , while the rhs . is for full qcd and we compare results in the confining ( low @xmath11 ) and the deconfining ( high @xmath11 ) phase . histograms for the distribution of the local modulus @xmath39 . we compare quenched ( lhs . ) and full qcd ( rhs . ) at low and high @xmath11 . the full curve is the haar measure distribution.,title="fig:",height=171 ] histograms for the distribution of the local modulus @xmath39 . we compare quenched ( lhs . ) and full qcd ( rhs . ) at low and high @xmath11 . the full curve is the haar measure distribution.,title="fig:",height=171 ] histograms for the distribution of the local phase @xmath40 . we compare quenched ( lhs . ) and full qcd ( rhs . ) at low and high @xmath11 . the full curve is the haar measure distribution.,title="fig:",height=264 ] histograms for the distribution of the local phase @xmath40 . we compare quenched ( lhs . ) and full qcd ( rhs . ) at low and high @xmath11 . the full curve is the haar measure distribution.,title="fig:",height=264 ] fig . 1 shows that the distribution of the modulus is essentially independent of @xmath11 , and furthermore is the same for both quenched and full qcd . the distribution of the modulus almost perfectly follows the haar measure distribution ( full curves in fig . 1 ) , @xmath41 = \int\!du \delta(\rho -\|\mbox{tr}\ , u \| ) $ ] , where @xmath42 is the su(3 ) haar measure . comparing the distributions below and above the transition / crossover temperature clearly shows that the local modulus is not involved in the transition , which in the case of pure su(3 ) gauge theory is even manifest as a first order jump of @xmath16 . obviously it must be the local phases @xmath40 which drive the transition . the histograms for the local phase in fig . 2 show a pronounced peak structure , with maxima at the three center angles 0 , @xmath43 and @xmath44 . in the confining phase ( top plots ) for all three maxima have the same height and we again observe no difference between the quenched and the dynamical case , both of which can be described by the haar measure distribution @xmath45 = \int\!du \delta(\varphi - \mbox{arg tr } \ , u)$ ] ( full curves in the top plots ) . in the deconfined phase ( bottom plots ) the situation is different . one of the three center phases is more populated . for full qcd , where the fermion determinant acts like an external field , it is always the real center sector ( phase values near 0 ) which becomes enhanced . in the quenched case any of the three center sectors may be selected spontaneously ( similar to , e.g. , the ising system where the magnetization has two signs to choose from in the symmetry broken phase ) . in the quenched distribution we show here we have chosen configurations where the phase angle of the summed polyakov loop @xmath6 is near @xmath46 and obviously the local phases are enhanced for that value . if one switches to any of the other equivalent center sectors , the distribution is shifted by @xmath47 . in the previous section we have seen that the transition to the deconfined phase is accompanied by an increasing population of the histograms for the phase @xmath40 of the local polyakov loop @xmath1 near one of the center angles . this accumulation of the local phases in one sector drives the increase of the expectation value @xmath16 , while the modulus of the local loops plays no role . it is interesting to note that at high temperatures ( bottom plots in fig . 2 ) there are still pronounced peaks also in the subdominant sectors . the question is whether the phase values at different positions @xmath3 are distributed independently , or if there are spatial domains where the phases of the local polyakov loops tend to align in the same center sector . the latter case is what is suggested by the effective action ( [ effact ] ) , where the first term favors parallel spins . in order to study the formation of domains we assign sector numbers @xmath48 to the sites @xmath3 , @xmath49 \ ; , \\ 0 & \ ; \mbox{for } \;\;\;\ ; \varphi(\vec{x } ) \ , \in \ , [ \,-\pi/3 + \delta \ , , \ , \pi/3 - \delta \ , ] \ ; , \\ + 1 & \ ; \mbox{for } \;\;\;\ ; \varphi(\vec{x } ) \ , \in \ , [ \,\pi/3 + \delta \ , , \ , \pi - \delta \ , ] \ ; . \end{array } \right.\ ] ] here @xmath50 is a free real and positive parameter which allows to cut lattice points @xmath3 where the corresponding phase @xmath40 is near a minimum of the distributions in fig . 2 . these points do not have a clear preference for one of the center sectors and the parameter @xmath50 allows one to remove them from the cluster analysis . the remaining lattice points @xmath3 can now be organized in clusters according to the sector numbers : we put two neighboring points @xmath51 into the same cluster if @xmath52 . for illustration purposes in fig . 3 we show the largest cluster for two quenched configurations , one below @xmath12 ( lhs . ) , the other one above ( rhs . ) . the plot is for lattice size @xmath53 and a value of @xmath50 chosen such that 39 % of the lattices points are cut . it is obvious , that above @xmath12 the largest cluster percolates ( stretches over all of the lattice ) , while it is finite below @xmath12 . of course the cutoff parameter @xmath50 will influence the size of the clusters , since with increasing @xmath50 less sites are available . we stress at this point that also for the characterization of the magnetic transitions in spin systems a similar reduction step is necessary in the construction of the clusters . this may be a reduced linking probability between neighbors with equal spins @xcite , but also more general approaches similar to the one used here were considered @xcite . as a consistency check of our cluster construction one may show @xcite that the points that survive the cut carry most of the signal of a rising polyakov loop above the transition / crossover temperature . largest clusters for two quenched configurations below @xmath12 ( lhs . ) and above ( rhs.).,title="fig:",width=264 ] largest clusters for two quenched configurations below @xmath12 ( lhs . ) and above ( rhs.).,title="fig:",width=264 ] weight @xmath54 of the largest cluster normalized with the volume @xmath8 as function of @xmath11.,title="fig:",height=200 ] weight @xmath54 of the largest cluster normalized with the volume @xmath8 as function of @xmath11.,title="fig:",height=200 ] in order to quantify the dependence of the cluster size on the temperature , in fig . 4 we show the expectation value of the weight @xmath54 ( i.e. , the number of lattice points ) of the largest cluster normalized by the total number @xmath8 of sites as a function of temperature . again the lhs . is for the quenched case ( at a cut of 39% for all volumes and temperatures ) , while the rhs . is for full qcd ( 19 % cut ) , and we compare three ( two ) different spatial volumes . both in the quenched and the dynamical case we observe that the clusters start out small below @xmath12 and grow quickly in size at the transition temperature . if one analyzes the percolation probability one finds that for the quenched case , where we have a true phase transition , this probability indeed approaches a step function as the volume is increased . for the dynamical case , where one observes only a crossover @xcite , the situation concerning percolation is not yet clear and further studies on several volumes will be necessary to settle this issue . we have shown that the phases @xmath40 of the local polyakov loops @xmath1 have preferred values near the center angles , and that neighboring sites have a tendency towards aligning these phases . the corresponding clusters were found to grow quickly near the transition / crossover temperature , and at least for the quenched case the deconfinement transition may be characterized by the onset of percolation for suitably defined clusters . so far this picture is established only for a fixed lattice spacing @xmath55 and some arbitrarily chosen value of the cutoff parameter @xmath50 which enters our cluster definition . if one wants to assign a physical significance to the center clusters , a way of constructing them such that a continuum limit can be taken is necessary . in particular this involves a prescription for connecting the scale @xmath55 in physical units to the clusters . we begin with defining the linear extension ( radius ) of a cluster by considering the expectation value of two - point correlators @xmath56 for sites @xmath57 within the same cluster . these 2-point functions decay exponentially , @xmath58 , and we use the parameter @xmath59 to define the radius of the cluster in units of the lattice spacing @xmath55 . using the value of the lattice constant @xmath55 in fm , the diameter of the clusters in physical units ( fm ) is then given by @xmath60 . the result will depend on both , the lattice spacing @xmath55 and the cutoff @xmath50 . in order to compare the physical size of the clusters for different lattice resolutions @xmath55 , we always adjust @xmath50 such that at a low temperature ( @xmath61 for the quenched case where this analysis is done ) we fix the physical diameter to a typical hadronic size , e.g. , @xmath62 fm . the corresponding value of @xmath50 is then kept fixed for all other temperatures , and is used to study @xmath63 as a function of @xmath11 . in fig . 5 we show the result for the cluster diameter @xmath63 in physical units as a function of temperature for the quenched case , comparing two different resolution scales @xmath55 ( i.e. , @xmath64 and @xmath65 ) . it is obvious that also in physical units the phase transition is signaled by a sudden increase of the cluster size . furthermore , the results for the two different scales fall onto a universal curve , which suggests that a continuum limit for the cluster size might exist . this question is analyzed in detail in a future publication @xcite , where we study the flow of the cutoff @xmath50 as a function of the resolution scale @xmath55 , arguing that the center clusters indeed have a continuum limit . * acknowledgments : * we thank christian lang and axel maas for interesting discussions . this work was partly supported by the dfg sfb tr 55 and the fwf dk 1203 . average cluster diameter in fm versus @xmath11 for two resolution scales ( quenched case).,width=366 ]	we analyze properties of local polyakov loops using quenched as well as dynamical su(3 ) gauge configurations for a wide range of temperatures . it is demonstrated that for both , the confined and the deconfined regime , the local polyakov loop prefers phase values near the center elements @xmath0 . we divide the lattice sites into three sectors according to these phases and show that the sectors give rise to the formation of clusters . for a suitable definition of these clusters we find that in the quenched case deconfinement manifests itself as the onset of percolation of the clusters . a possible continuum limit of the center clusters is discussed .
ngc 5728 is a well known prototype of embedded bars . it is classified as ( r1)sab(r)a , and a seyfert 2 ( distance 33 mpc 1=180 pc ) . its nuclear bar is oriented at 45 degrees of the main bar ( shaw et al 1993 ) , and could rotate faster . petitpas & wilson ( 2002 ) have obtained a co map of ngc 5728 with the ovro interferometer . they found a total co mass of 3.1 10@xmath3 m@xmath2 . however , they underestimate the extended component , filtered out by the interferometer , and miss a significant part of the co emission , due to restricted velocity coverage . we report here co(1 - 0 ) and co(2 - 1 ) observations with the iram-30 m telescope , with 23 and 12 " beams . the molecular gas distribution does not follow the nuclear bar in the center , but is strongly asymmetric . it follows the ionization cone , with the geometry found by wilson et al ( 1993 ) . at larger scale , there is a tendency for the co intensity to follow the primary bar , along its characteristic dust lanes ( cf fig . the total h@xmath0 mass ( without including the he ) is 3.1 10@xmath1 m@xmath2 with a conversion factor of 3 10@xmath4 cm-2/(k.km / s ) . it is larger than the hi mass of 2.5 10@xmath1 m@xmath2 . at the tips of the bar , strong co(1 - 0 ) emission is found , associated with star formation , due to orbit crowding ( in particular on the receding side , at 35 " @xmath5 7kpc ) . the total molecular gas mass is about 10 % of the stellar disc mass . the co(2 - 1)/co(1 - 0 ) intensity ratio is in general quite low ( between 0.5 and 1 ) . one interesting feature is the presence of two strips of higher co ratio parallel to the radio jet direction . it could indicate an enhancement of the co excitation along the wall of the ionization cone .	we present co(1 - 0 ) and co(2 - 1 ) maps of the seyfert galaxy ngc 5728 . although a stellar nuclear bar structure is clearly identified in the near - infrared images in the central 10 `` , inside the ring identified as the ilr of the primary bar , there is no nuclear bar structure in the molecular gas . instead , the co emission reveals an elongated structure , 15 '' in length , beginning at the nucleus ( defined by the radio center ) aligned with the jet / ionisation cone , at a pa of 127 degrees . this morphology , not frequently observed in seyfert galaxies , may be interpreted in terms of enhanced co excitation along the walls of the cone . kinematical perturbations along the cone support this scenario . at larger - scale , co emission is tracing the primary bar , and outer ring structure . the total molecular mass , estimated from the co emission , is m(h@xmath0 ) = 3.1 10@xmath1 m@xmath2 .
perturbation theory is a powerful tool to obtain reliable predictions for physical observables within the standard model of particle physics or its extensions . due to the high precision of experimental results , e.g. , at the cern large hadron collider ( lhc ) or at the @xmath2 factories , it is on the one hand mandatory to advance the development of tools , which can be used for higher order calculations . on the other hand it is necessary to improve the understanding of the perturbative structure of quantum field theories . form factors are ideal objects to obtain deeper insight into the latter . especially in the context of qcd they are indispensable tools to investigate the infrared structure of scattering amplitudes to high orders in perturbation theory . moreover , from the pole part it is possible to extract universal process - independent quantities @xcite like the cusp anomalous dimension which can be extracted from the @xmath3 pole of the form factor . the finite parts of the form factors serve as building blocks for a variety of physical processes . for example , the quark - anti - quark - photon form factor enters the virtual corrections of the drell - yan process for the production of lepton pairs at hadron colliders . in this paper we consider the quark - anti - quark - photon form factor which is conveniently obtained from the photon - quark vertex function @xmath4 by applying an appropriate projector . in @xmath5 space - time dimensions we have @xmath6 with @xmath7 where @xmath8 and @xmath9 are the incoming quark and anti - quark momenta and @xmath10 is the momentum of the photon . we perform our calculation in the framework of qcd keeping the number of colours , @xmath1 , generic . in the limit of large @xmath1 the calculation of @xmath0 is simplified since only planar feynman diagrams contribute . this is the limit we consider in this paper . besides @xmath1 we also keep the number of active quark flavours , @xmath11 as a parameter and thus have at four - loop order the colour structures @xmath12 , @xmath13 , @xmath14 , @xmath15 where each factor of @xmath11 counts the number of closed fermion loops . two- and three - loop corrections to @xmath0 have been computed in refs . @xcite . to obtain the four - loop corrections two obstacles need to be overcome : ( i ) the reduction to a set of basis integrals and ( ii ) the ( if possible ) analytic calculation of the latter . recently , the first steps towards four loops have been initiated by computing the fermionic contributions to @xmath0 in the planar limit @xcite . the @xmath16 terms have been confirmed in ref . @xcite using different methods both for the reduction and the computation of the master integrals . let us mention that the four - loop corrections to the cusp anomalous dimension with two and three closed fermion loops have also been obtained in ref . @xcite . in ref . @xcite one finds a discussion of non - planar master intergrals relevant for the four - loop form factor in a @xmath17 supersymmetric yang - mills theory . in the present paper , we evaluate the @xmath18 contribution and therefore complete the evaluation of the form factors and anomalous dimensions in the limit of large @xmath1 . in the next section we provide some technical details , in particular to the calculation of the most complicated master integral , and we discuss our results in section [ sec::res ] . we provide explicit expressions for the four - loop cusp and collinear anomalous dimensions in the planar limit . furthermore , we provide results for the finite part of @xmath19 . sample four - loop feynman diagrams contributing to @xmath0 . solid , curly and wavy lines represent quarks , gluons and photons , respectively . all particles are massless.,title="fig:",scaledwidth=30.0% ] sample four - loop feynman diagrams contributing to @xmath0 . solid , curly and wavy lines represent quarks , gluons and photons , respectively . all particles are massless.,title="fig:",scaledwidth=30.0% ] sample four - loop feynman diagrams contributing to @xmath0 . solid , curly and wavy lines represent quarks , gluons and photons , respectively . all particles are massless.,title="fig:",scaledwidth=30.0% ] for the calculation of @xmath0 we use a well - tested automated chain of programs which work hand - in - hand . the feynman amplitudes are generated with qgraf @xcite . since there is no possibility to select already at this point the planar diagrams also non - planar amplitudes are generated and we obtain in total 1 , 15 , 337 and 9784 diagrams at one , two , three and four loops . sample feynman diagrams at four loops can be found in fig . [ fig::diags ] . next , we transform the output to form @xcite notation using q2e and exp @xcite . the program exp furthermore maps each feynman diagram to predefined integral families for massless four - loop vertices with two different non - vanishing external momenta ; 68 of them are of planar type . at this point we perform the dirac algebra and decompose the numerator into terms which appear in the denominator . this allows us to express each feynman integral as a linear combination of scalar functions which belong to the corresponding family . after exploiting the symmetries connected to the exchange of the external momenta we can reduce the number of families , for which integral tables have to be generated , from 68 to 38 . note that for the fermionic contributions , which have been considered in ref . @xcite , only 24 families are needed . for the reduction to master integrals we use the program fire @xcite which we apply in combination with litered @xcite . we observe that the non - fermionic diagrams lead to more complex integrals for which the reduction time significantly increases . let us remark that we have adopted feynman gauge for the calculation of the non - fermionic parts . once reduction tables for each family are available we apply tsort , which is part of the latest fire version @xcite . it is based on ideas presented in ref . @xcite to establish relations between between primary master integrals and thus minimize their number . in this way we arrive at 99 master integrals . in the remainder of this section we describe in detail the calculation of the most complicated integral corresponding to the graph of fig . [ fig::i99 ] , @xmath20 . massless four - loop vertex diagram with two external ( incoming ) momenta @xmath10 and @xmath21 with @xmath22 and @xmath23 . the master integral @xmath20 corresponds to @xmath24.,scaledwidth=50.0% ] to have the possibility to apply differential equations for the evaluation of the form factor master integral corresponding to the graph of fig . [ fig::i99 ] we introduce a second mass scale by considering @xmath25 and derive differential equations with respect to the ratio of the two scales , @xmath26 . this strategy was advocated in @xcite and used for the four - loop form - factor integrals in @xcite . we derive differential equations for the corresponding 332 master integrals using litered @xcite . to solve our differential equations we use an important observation made in ref . it has been suggested to turn from the basis of primary master integrals to a so - called canonical basis where the corresponding integrals satisfy a system differential equations which has a particular structure : the dependence on @xmath27 appears as a linear prefactor and the matrix in front of the vector of master integrals has only simple poles in @xmath26 , i.e. has only so - called fuchsian singularities . such a system can then easily be solved in terms of iterated integrals . in ref . @xcite , it was proposed that choosing integrals with constant leading singularities provides a canonical basis of the differential equations . in subsequent work , this was used in a variety of cases , see e.g. refs . @xcite . as was motivated in ref . @xcite , this choice can be done in a systematic way at the level of the loop integrand . the first algorithm to convert a given differential system to a canonical form at the level of the differential equations has been provided in the case of one variable in ref . @xcite ( therein called @xmath27-form , see also ref . recently a public implementation @xcite of the algorithm of ref . @xcite in a computer code fuchsia became available . in this paper we follow @xcite to construct a canonical basis . an independent evaluation of @xmath20 can be found in ref . @xcite where the canonical basis was chosen at the level of the loop integrand . the size of the system is large but , as it concerns the diagonal blocks , their size is at most @xmath28 . consequently , the reduction of diagonal blocks within the approach of ref . @xcite is simple . note that already after this step one might claim that the solution is expressible in terms of harmonic polylogarithms ( and also construct this solution ) . nevertheless , we follow the prescription of section 7 of ref . @xcite to reduce the whole system to an @xmath27-form . similar to ref . @xcite we find that the differential equation - based hierarchy of the set of the master integrals is too restrictive and the use of a sector - based hierarchy when factoring @xmath27 out of the whole matrix is necessary . the family of one - scale feynman integrals ( with @xmath24 and with @xmath29 ) corresponding to fig . [ fig::i99 ] contains 76 master integrals . after introducing @xmath30 we define @xmath31 and obtain a family of feynman integrals with 332 master integrals . our strategy is to turn from a primary basis to a canonical basis , solve differential equations for the canonical basis , evaluate the naive values of the elements of the canonical basis at @xmath32 ( i.e. , setting @xmath32 under the integral sign ) from which it will be straightforward to obtain analytical results for the primary master integrals of our one - scale family , in particular , @xmath20 . following ref . @xcite we arrive at a canonical basis @xmath33 which is obtained from the primary basis @xmath34 by a linear transformation with a matrix @xmath35 , @xmath36 the vectors @xmath34 and @xmath33 have @xmath37 entries and @xmath35 is a @xmath38 matrix . the dependence of @xmath34 , @xmath33 and @xmath35 on @xmath26 and @xmath27 has been suppressed . it is convenient @xcite to normalize the canonical master integrals such that they have uniformly transcendental @xmath27-expansion which starts from @xmath39 . in our four - loop case , one has to compute expansion terms including @xmath40 terms and we have @xmath41 the canonical basis @xmath33 satisfies ( by definition ) a differential equation of the form @xmath42 where @xmath43 with constant matrices @xmath44 and @xmath45 . it is straightforward to construct the generic solution of ( [ eq::dgl ] ) order - by - order in @xmath27 in terms of harmonic polylogarithms ( hpls ) @xcite with letters 0 and 1 , with @xmath46 unknown constants . to fix these constants we use boundary conditions at the point @xmath47 . the primary master integrals are regular at this point where the integrals become propagator - type integrals which are well known @xcite . in particular , all the corresponding 28 master integrals are known analytically @xcite in an @xmath27-expansion up to weight 12 and have been cross checked numerically @xcite . we obtain the boundary values of the elements of the canonical basis @xmath33 at @xmath47 by inverting eq . ( [ eq::t ] ) and considering the limit @xmath48 . the matrix @xmath49 involves elements which develop poles up to order @xmath50 . this requires that the first seven expansion terms for @xmath48 of the primary master integrals @xmath34 have to be computed . the corresponding reduction tables are again generated with fire . note that the resulting two - point integrals with external momentum @xmath10 have scalar products in the numerator involving the momentum @xmath21 which complicates the calculation . it is an important cross check of the calculation that all poles in @xmath51 cancel in the combination @xmath52 and we obtain the values for @xmath33 at @xmath47 , in an @xmath27-expansion up to @xmath40 . this fixes the solution of the system of differential equations in eq . ( [ eq::dgl ] ) . in a next step we analyze the leading asymptotic behaviour of @xmath33 near @xmath32 . on the one hand , we obtain it from the differential equations ( [ eq::dgl ] ) where the term @xmath53 on the right - hand side can be neglected and the solution has the form @xmath54 . the quantity @xmath55 can be evaluated using the mathematica command matrixexp [ ] which leads to a @xmath38 matrix where each element is a linear combination of terms @xmath56 with integer @xmath57 . in general both non - positive @xmath57 and positive @xmath57 might appear . however , in the case of feynman integrals only terms with non - positive @xmath57 can be present which we use as a check . on the other hand , the leading asymptotic behaviour in the limit @xmath58 can also be obtained with the help of the mathematica package hpl @xcite from our analytic expression for the canonical basis . matching the two expressions provides values for the vector @xmath59 in an @xmath27-expansion up to @xmath40 and terms @xmath56 with @xmath60 . note that this step involves powers of @xmath61 terms ; their cancellation in the matching provides a welcome check for our calculation . finally , the naive value for @xmath62 at @xmath32 is obtained by setting all terms @xmath56 with @xmath63 to zero in the expression for @xmath64 . in the last step , using eq . ( [ eq::t ] ) , we compute the naive values of the elements of the primary basis @xmath34 from the naive expansion of the canonical basis @xmath65 near @xmath58 . note that some of the matrix elements of @xmath35 involve singularities up to order @xmath66 . thus , the naive expansion of @xmath65 up to order @xmath67 is needed . it is obtained following the prescription outlined in ref . @xcite where the expansion terms can be computed from the leading order asymptotics at @xmath68 after recursively solving matrix equations with @xmath38 entries . after inserting the expansion of @xmath33 in eq . ( [ eq::t ] ) the poles cancel and the naive values of the primary master integrals at @xmath32 are obtained . the naive value of one of the elements of our primary basis @xmath34 is nothing but the one - scale master integral @xmath20 ( cf . [ fig::i99 ] ) for which we obtain the following analytic result @xmath69 + \frac{1}{\epsilon^6 } \bigg [ \frac{13}{576 } \bigg ] + \frac{1}{\epsilon^5 } \bigg [ -\frac{101}{576}-\frac{\pi ^2}{48 } \bigg ] \nonumber\\&&\mbox { } + \frac{1}{\epsilon^4 } \bigg [ -\frac{17 \zeta _ 3}{54}+\frac{5 \pi ^2}{36}+\frac{145}{96 } \bigg ] + \frac{1}{\epsilon^3 } \bigg [ \frac{1775 \zeta _ 3}{432}-\frac{767 \pi ^4}{17280}-\frac{5 \pi ^2}{8}-\frac{1669}{144 } \bigg ] \nonumber\\&&\mbox { } + \frac{1}{\epsilon^2 } \bigg [ -\frac{83}{72 } \pi ^2 \zeta _ 3-\frac{21899 \zeta _ 3}{864}-\frac{3659 \zeta _ 5}{360}+\frac{31333 \pi ^4}{103680}+\frac{659 \pi ^2}{288}+\frac{11243}{144 } \bigg ] \nonumber\\&&\mbox { } + \frac{1}{\epsilon } \bigg [ -\frac{40231 \zeta _ 3 ^ 2}{1296}+\frac{745 \pi ^2 \zeta _ 3}{288}+\frac{18751 \zeta _ 3}{144}+\frac{50191 \zeta _ 5}{360}-\frac{277703 \pi ^6}{2177280}-\frac{14015 \pi ^4}{10368 } \nonumber\\&&\mbox { } -\frac{149 \pi ^2}{24}-\frac{22757}{48 } \bigg ] \nonumber\\&&\mbox { } + \bigg [ \frac{39173 \zeta _ 3 ^ 2}{324}-\frac{77399 \pi ^4 \zeta _ 3}{25920}+\frac{4013 \pi ^2 \zeta _ 3}{432}-\frac{259559 \zeta _ 3}{432}-\frac{568 \pi ^2 \zeta _ 5}{45}-\frac{1123223 \zeta _ 5}{1440 } \nonumber\\&&\mbox { } -\frac{2778103 \zeta _ 7}{4032}+\frac{3129533 \pi ^6}{4354560}+\frac{28201 \pi ^4}{5760}+\frac{173 \pi ^2}{36}+\frac{382375}{144 } \bigg ] \nonumber\\&&\mbox { } + \epsilon \bigg [ \frac{4931 s_{8a}}{30}+\frac{2615}{144 } \pi ^2 \zeta _ 3 ^ 2-\frac{276671 \zeta _ 3 ^ 2}{2592}-\frac{2702413 \zeta _ 5 \zeta _ 3}{1080}+\frac{154037 \pi ^4 \zeta _ 3}{31104 } \nonumber\\&&\mbox { } -\frac{55327 \pi ^2 \zeta _ 3}{432}+\frac{1100461 \zeta _ 3}{432}+\frac{205 \pi ^2 \zeta _ 5}{9}+\frac{155029 \zeta _ 5}{48}+\frac{2732549 \zeta _ 7}{1008}-\frac{665217829 \pi ^8}{1306368000 } \nonumber\\&&\mbox { } -\frac{131003 \pi ^6}{45360}-\frac{747929 \pi ^4}{51840}+\frac{2995 \pi ^2}{36}-\frac{2005247}{144 } \bigg ] \bigg\}\,,\end{aligned}\ ] ] where @xmath70 is riemann s zeta function evaluated at @xmath71 and @xmath72 @xmath73 are multiple zeta values given by @xmath74 as by - product we also obtain analytic results for the remaining 75 one - scale master integrals and we find agreement with the results obtained in ref . this constitutes a further cross check for our procedure . we want to stress that the calculation which is outlined in this section is largely independent from the one performed in ref . @xcite . this section is devoted to the analytic results of the cusp and collinear anomalous dimensions and the finite part of @xmath0 . generic formulae where the pole part of @xmath0 is parametrized in terms of the cusp and collinear anomalous dimensions and the qcd beta function can , e.g. , be found in refs . in what follows we use eq . ( 2.3 ) of ref . @xcite which displays the pole parts of @xmath19 up to four - loop order . in this formula it is assumed that the one - loop coefficient of the beta function is given by @xmath75 with @xmath11 being the number of active quarks and the coefficients of the anomalous dimensions are defined through @xmath76 with @xmath77 and @xmath78 is the renormalized coupling constant with @xmath11 active flavours . from eq . ( 2.3 ) of ref . @xcite one observes that the four - loop corrections of @xmath79 follows from the @xmath3 term of @xmath19 and @xmath80 from the linear pole terms . in the following we start with explicit results for the cusp and collinear anomalous dimensions . the four - loop corrections to @xmath79 reads @xmath81 we note that , after taking into account the difference between fundamental and adjoint representation of the external fields , the leading transcendental piece of this expression agrees with the result for planar @xmath17 super yang - mills @xcite , in agreement with expectations from @xcite . note that @xmath82 entering the @xmath3 pole of @xmath19 is multiplied by @xmath83 . for this reason there is only a @xmath84 factor in front of the @xmath11-independent term in eq . ( [ eq::gamma_cusp3 ] ) . the one- , two- and three - loop corrections in the large-@xmath1 limit can be found in eq . ( 2.6 ) of ref . @xcite where also the fermionic part of @xmath82 is shown . the four - loop coefficient of the collinear anomalous dimension is given by @xmath85+\left(-\frac{8 \pi ^4}{1215 } -\frac{356 \zeta _ 3}{243}-\frac{2 \pi ^2}{81}+\frac{18691}{13122}\right ) n_c n_f^3+\left(-\frac{2}{3 } \pi ^2 \zeta _ 3 \right.\nonumber\\&&\left.\mbox { } + \frac{166 \zeta _ 5}{9}+\frac{331 \pi ^4}{2430 } -\frac{2131 \zeta _ 3}{243 } -\frac{68201 \pi ^2}{17496}-\frac{82181}{69984}\right ) n_c^2 n_f^2 \nonumber\\&&\mbox { } + n_c^4 \left(\frac{1175 \zeta _ 3 ^ 2}{9}+\frac{82 \pi ^4 \zeta _ 3}{45}-\frac{377 \pi ^2 \zeta _ 3}{6}+\frac{867397 \zeta _ 3}{972}+24 \pi ^2 \zeta _ 5 - 1489 \zeta _ 5 + 705 \zeta _ 7 \right.\nonumber\\&&\left.\mbox { } + \frac{114967 \pi ^6}{204120}-\frac{59509 \pi ^4}{9720}-\frac{120659 \pi ^2}{17496}-\frac{187905439}{839808}\right ) \ , . \label{eq::gamma_q}\end{aligned}\ ] ] the one- , two- and three - loop corrections and the fermionic four - loop terms to @xmath86 are listed in eq . ( 2.7 ) of ref . the @xmath12 term is new . finally , we also present the finite part of the form factor . we parametrize the perturbative expansion in terms of the renormalized coupling constant and set @xmath87 . furthermore , it is convenient to consider @xmath19 which leads to more compact expressions . thus , we have the following parametrization @xmath88 in the large-@xmath1 limit the four - loop term reads @xmath89 where the ellipses refer to the fermionic contributions which are given in eq . ( 2.8 ) of ref . @xcite . for convenience of the reader we provide the results for the form factor @xmath0 expanded in the bare strong coupling constant in an ancillary file which can be downloaded from ` https://www.ttp.kit.edu/preprints/2016/ttp16-055/ ` . this file also contains the lower - loop results expanded to higher order in @xmath27 . furthermore , it contains the dependence of the renormalization scale @xmath90 . we compute the photon - quark form factor to four loops up to the finite term in @xmath27 in the large-@xmath1 limit which is obtained from the planar feynman diagrams . from the pole parts we extract the cusp and collinear anomalous dimensions . we discuss in detail the calculation of the most complicated master integral ( see fig . [ fig::i99 ] ) and present analytic results expanded in @xmath27 up to transcendental weight eight . an independent calculation of this integral , together with a discussion of the remaining master integrals can be found in refs . @xcite to the required order in @xmath27 . we want to remark that the same master integrals enter the higgs gluon form factor in the planar limit . however , the corresponding reduction is significantly more complicated . the logical next step is the calculation of the non - planar contributions . note that also here both the reduction and the computation of the master integrals turn out to be much more complicated . we are grateful to gang yang for checking numerically the highest poles of our result for @xmath20 . is supported in part by a gfk fellowship and by the prisma cluster of excellence at mainz university . this work is supported by the deutsche forschungsgemeinschaft through the project `` infrared and threshold effects in qcd '' . * note added * : + after submission of our mansucript to arxiv.org we were contacted by colleagues drawing our attention to a talk by ben ruijl at the university of zurich where the result for @xmath82 has been shown . the result agrees with our eq . ( 11 ) . a. h. mueller , phys . d * 20 * ( 1979 ) 2037 . doi:10.1103/physrevd.20.2037 j. c. collins , phys . rev . d * 22 * ( 1980 ) 1478 . doi:10.1103/physrevd.22.1478 a. sen , phys . d * 24 * ( 1981 ) 3281 . doi:10.1103/physrevd.24.3281 l. magnea and g. f. sterman , phys . rev . d * 42 * ( 1990 ) 4222 . doi:10.1103/physrevd.42.4222 g. p. korchemsky and a. v. radyushkin , phys . b * 279 * ( 1992 ) 359 doi:10.1016/0370 - 2693(92)90405-s [ hep - ph/9203222 ] . i. a. korchemskaya and g. p. korchemsky , phys . b * 287 * ( 1992 ) 169 . doi:10.1016/0370 - 2693(92)91895-g g. f. sterman and m. e. tejeda - yeomans , phys . b * 552 * ( 2003 ) 48 doi:10.1016/s0370 - 2693(02)03100 - 3 [ hep - ph/0210130 ] . g. kramer and b. lampe , z. phys . c * 34 * ( 1987 ) 497 [ erratum - ibid . c * 42 * ( 1989 ) 504 ] . t. matsuura and w. l. van neerven , z. phys . c * 38 * ( 1988 ) 623 . t. matsuura , s. c. van der marck and w. l. van neerven , nucl . b * 319 * ( 1989 ) 570 . t. gehrmann , t. huber and d. maitre , phys . b * 622 * ( 2005 ) 295 [ arxiv : hep - ph/0507061 ] . a. baikov , k. g. chetyrkin , a. v. smirnov , v. a. smirnov and m. steinhauser , phys . * 102 * ( 2009 ) 212002 doi:10.1103/physrevlett.102.212002 [ arxiv:0902.3519 [ hep - ph ] ] . t. gehrmann , e. w. n. glover , t. huber , n. ikizlerli and c. studerus , jhep * 1006 * ( 2010 ) 094 doi:10.1007/jhep06(2010)094 [ arxiv:1004.3653 [ hep - ph ] ] . r. n. lee and v. a. smirnov , jhep * 1102 * ( 2011 ) 102 doi:10.1007/jhep02(2011)102 [ arxiv:1010.1334 [ hep - ph ] ] . t. gehrmann , e. w. n. glover , t. huber , n. ikizlerli and c. studerus , jhep * 1011 * ( 2010 ) 102 doi:10.1007/jhep11(2010)102 [ arxiv:1010.4478 [ hep - ph ] ] . j. m. henn , a. v. smirnov , v. a. smirnov and m. steinhauser , jhep * 1605 * ( 2016 ) 066 doi:10.1007/jhep05(2016)066 [ arxiv:1604.03126 [ hep - ph ] ] . a. von manteuffel and r. m. schabinger , arxiv:1611.00795 [ hep - ph ] . j. davies , a. vogt , b. ruijl , t. ueda and j. a. m. vermaseren , arxiv:1610.07477 [ hep - ph ] . r. boels , b. a. kniehl and g. yang , nucl . b * 902 * ( 2016 ) 387 doi:10.1016/j.nuclphysb.2015.11.016 [ arxiv:1508.03717 [ hep - th ] ] . p. nogueira , j. comput . phys . * 105 * ( 1993 ) 279 . j. kuipers , t. ueda , j. a. m. vermaseren and j. vollinga , comput . phys . commun . * 184 * ( 2013 ) 1453 doi:10.1016/j.cpc.2012.12.028 [ arxiv:1203.6543 [ cs.sc ] ] . r. harlander , t. seidensticker and m. steinhauser , phys . b * 426 * ( 1998 ) 125 [ hep - ph/9712228 ] . t. seidensticker , hep - ph/9905298 . a. v. smirnov , jhep * 0810 * ( 2008 ) 107 doi:10.1088/1126 - 6708/2008/10/107 [ arxiv:0807.3243 [ hep - ph ] ] . a. v. smirnov and v. a. smirnov , comput . commun . * 184 * ( 2013 ) 2820 doi:10.1016/j.cpc.2013.06.016 [ arxiv:1302.5885 [ hep - ph ] ] . a. v. smirnov , comput . commun . * 189 * ( 2015 ) 182 doi:10.1016/j.cpc.2014.11.024 [ arxiv:1408.2372 [ hep - ph ] ] . r. n. lee , arxiv:1212.2685 [ hep - ph ] . r. n. lee , j. phys . * 523 * ( 2014 ) 012059 doi:10.1088/1742 - 6596/523/1/012059 [ arxiv:1310.1145 [ hep - ph ] ] . j. m. henn , a. v. smirnov and v. a. smirnov , jhep * 1403 * ( 2014 ) 088 doi:10.1007/jhep03(2014)088 [ arxiv:1312.2588 [ hep - th ] ] . j. m. henn , a. v. smirnov and v. a. smirnov , to appear . j. m. henn , phys . * 110 * ( 2013 ) 251601 doi:10.1103/physrevlett.110.251601 [ arxiv:1304.1806 [ hep - th ] ] . j. m. henn , j. phys . a * 48 * ( 2015 ) 153001 doi:10.1088/1751 - 8113/48/15/153001 [ arxiv:1412.2296 [ hep - ph ] ] . t. gehrmann , a. von manteuffel , l. tancredi and e. weihs , jhep * 1406 * ( 2014 ) 032 doi:10.1007/jhep06(2014)032 [ arxiv:1404.4853 [ hep - ph ] ] . p. mastrolia , m. argeri , s. di vita , e. mirabella , j. schlenk , u. schubert and l. tancredi , pos ll * 2014 * ( 2014 ) 007 . c. meyer , arxiv:1611.01087 [ hep - ph ] . r. n. lee , jhep * 1504 * ( 2015 ) 108 doi:10.1007/jhep04(2015)108 [ arxiv:1411.0911 [ hep - ph ] ] . o. gituliar and v. magerya , pos ll * 2016 * ( 2016 ) 030 [ arxiv:1607.00759 [ hep - ph ] ] . r. n. lee and v. a. smirnov , jhep * 1610 * ( 2016 ) 089 doi:10.1007/jhep10(2016)089 [ arxiv:1608.02605 [ hep - ph ] ] . e. remiddi and j. a. m. vermaseren , int . j. mod . phys . a * 15 * ( 2000 ) 725 doi:10.1142/s0217751x00000367 [ hep - ph/9905237 ] . p. a. baikov and k. g. chetyrkin , nucl . b * 837 * ( 2010 ) 186 doi:10.1016/j.nuclphysb.2010.05.004 [ arxiv:1004.1153 [ hep - ph ] ] . r. n. lee , a. v. smirnov and v. a. smirnov , nucl . b * 856 * ( 2012 ) 95 doi:10.1016/j.nuclphysb.2011.11.005 [ arxiv:1108.0732 [ hep - th ] ] . a. v. smirnov and m. tentyukov , nucl . b * 837 * ( 2010 ) 40 doi:10.1016/j.nuclphysb.2010.04.020 [ arxiv:1004.1149 [ hep - ph ] ] . d. maitre , comput . commun . * 174 * ( 2006 ) 222 [ arxiv : hep - ph/0507152 ] . w. wasow , `` asymptotic expansions for ordinary differential equations '' , interscience publishers john wiley & sons , inc . , new york - london - sydney , pure and applied mathematics , vol . xiv , 1965 . t. becher and m. neubert , jhep * 0906 * ( 2009 ) 081 erratum : [ jhep * 1311 * ( 2013 ) 024 ] doi:10.1088/1126 - 6708/2009/06/081 , 10.1007/jhep11(2013)024 [ arxiv:0903.1126 [ hep - ph ] ] . z. bern , m. czakon , l. j. dixon , d. a. kosower and v. a. smirnov , phys . d * 75 * ( 2007 ) 085010 doi:10.1103/physrevd.75.085010 [ hep - th/0610248 ] . n. beisert , b. eden and m. staudacher , j. stat . * 0701 * ( 2007 ) p01021 doi:10.1088/1742 - 5468/2007/01/p01021 [ hep - th/0610251 ] . a. v. kotikov , l. n. lipatov , a. i. onishchenko and v. n. velizhanin , phys . b * 595 * ( 2004 ) 521 erratum : [ phys . b * 632 * ( 2006 ) 754 ] doi:10.1016/j.physletb.2004.05.078 , 10.1016/j.physletb.2005.11.002 [ hep - th/0404092 ] .	we compute the four - loop qcd corrections to the massless quark - anti - quark - photon form factor @xmath0 in the large-@xmath1 limit . from the pole part we extract analytic expressions for the corresponding cusp and collinear anomalous dimensions . pacs numbers : 11.15.bt , 12.38.bx , 12.38.cy
the cast ( cern axion solar telescope ) experiment is using a decommissioned lhc dipole magnet to convert solar axions into detectable x - ray photons . axions are light pseudoscalar particles that arise in the context of the peccei - quinn@xcite solution to the strong cp problem and can be dark matter candidates@xcite . stars could produce axions via the primakoff conversion of the plasma photons . the cast experiment is pointing at our closest star , the sun , aiming to detect solar axions . the detection principle is based on the coupling of an incoming axion to a virtual photon provided by the transverse field of an intense dipole magnet , being transformed into a real , detectable photon that carries the energy and the momentum of the original axion . the axion to photon conversion probability is proportional to the square of the transverse field of the magnet and to the active length of the magnet . using an lhc magnet ( @xmath7 t and @xmath8 m long ) improves the sensitivity by a factor 100 compared to previous experiments . the cast experiment has been taking data since 2003 providing the most restrictive limits on the axion - photon coupling @xcite for masses @xmath9ev . at this mass the sensitivity is degraded due to coherence loss . in order to restore coherence , the magnet can be filled with a buffer gas providing an effective mass to the photon@xcite . by changing the pressure of the buffer gas in steps , one can scan an entire range of axion mass values . at the end of 2005 the cast experiment started such a program , entering its phase ii by filling the magnet bore with he gas . from 2005 to 2007 , the magnet bore was filled with @xmath10 gas extending our sensitivity to masses up to @xmath11ev , final results will be presented here . from march 2008 onwards the magnet bore has been filled with @xmath12 and the sensitivity should be increased to sensivities up to @xmath13ev by the end of the @xmath12 run in 2010 . the cast set up has been described elsewhere @xcite . from 2002 to 2006 three x - ray detectors were mounted on the two sides of the magnet : a conventional tpc@xcite covering both magnet bores looking for sunset axions ; in the sunrise side one of the bores was covered by a micromegas detector@xcite and in the other bore a ccd detector coupled to a telescope@xcite improving the signal to background ratio by a factor 150 . in 2006 the tpc started to show a degraded performance due to aging . it was then decided to replace the sunset tpc and the existing micromegas detector in the sunrise side by a new generation of micromegas detectors@xcite that coupled with suitable shielding would improve greatly their performance . the new detectors were commissioning end of 2007 and by mid 2008 they have already shown an improvement in performance that has been translated in a background reduction of a factor 15 compared to the tpc performances and a factor 3 compared to the standard micromegas detector used without shielding till 2006 . the ccd detector will also be upgraded by the fall 2009 by a new detector with improved performance : better low energy response , lower intrinsic background by using more radio - pure materials and less out of time events . in 2005 , the experiment went through a major upgrade to allow operation with he buffer gas in the cold bore . this upgrade was done in two steps : first the system was designed for operation using @xmath10 and in 2007 the system was upgraded for operation at higher buffer gas densities using @xmath12 . the system has been designed to control the injection of he in the magnet bores with precision and to monitor accurately the gas pressure and temperature@xcite . special care has been taken to achieve high precision in the reproducibility of steps ( @xmath14 mbar ) and to protect the system for @xmath12 loss . the @xmath12 system has been operating succesfully since december 2007 . as during phase i , the tracking data ( magnet pointing the sun ) represented about [email protected] hours per day while the rest of the day was used to measure background . the procedure was to daily increase the @xmath10 density so that sunrise and sunset detectors measure every pressure . every specific pressure of the gas allows to test a specific axion mass having a new discovery potential . the @xmath10 data recorded end of 2005 and 2006 represents around 300 hours of tracking data and 10 times more hours of background data for each detector , covering 160 pressure settings allowing to scan a new axion mass range between 0.02 and 0.39 ev . an independent analysis was performed for each data set of the three different detectors . a combined preliminary result was derived where from the absence of a signal above background cast excludes a new range in the @xmath16@xmath1 plane shown in figure [ fig : exclusion ] from axion masses of 0.02ev ( phase i ) up to masses of 0.39ev . this parameter space was not previously explored in laboratory experiments . cast has therefore entered the qcd axion band for the first time in this range of axion masses , excluding an important portion of the axion parameter space . the final results have been published in @xcite . the collaboration has performed by - product analysis of the data taken , to look for other axion scenario to which cast would also be sensitive . the tpc phase i data has been reanalysed in order look for 14 kev axions coming from m1 transitions . in addition , data taken with a calorimeter during the phase i , were used to search for high energy ( mev ) lines from high energy axion conversion @xcite . more recently a few days of data were taken with a visible detector coupled to one end of the cast magnet @xcite , in search for axions with energy in the `` visible '' range . it is foressen that a permanent setup will be installed in the experiment in order to take data without interfering with the standard program of cast . at present , we are running with @xmath12 since 2008 and by the end of this year we shoud have reached sensitivities of around @xmath170.8ev . part of phase ii ) is compared with constraints obtained from the sumico experiment ( the tokyo helioscope ) and hb stars . the red dashed line shows our prospects for the @xmath12 run started in march 2008 . the vertical line ( hdm ) is the hot dark matter limit for hadronic axions @xmath18 inferred from observations of the cosmological large - scale structure . the yellow band represents typical theoretical models with @xmath19 in the range 0.077 where the green solid line corresponds to the case when @xmath20 is assumed . limits from laser , microwave and underground detectors @xcite for axion searches have been included.,scaledwidth=60.0% ] cast original physics program will finish at the end of 2011 . the cast collaboration is looking into possibilities in order to achieve greater sensitivities . the sensitivity of helioscope axion searches depends strongly on the magnet s characteristics . ongoing r & d on dipoles will lead to stronger and bigger magnets in the coming years ( 2013 - 2015 ) . such magnets could either be adopted to the existing infrastructure of cast , thus eliminating the cost to the magnet itself , or be used with a new tracking and cryogenics system , improving thus many characteristics ( like daily tracking time , safety etc ) but increasing the cost . in parallel , effort will be devoted on the development of high efficiency focusing devices and new electronics for the detectors , aiming to achieve very low background levels . all this would allow pushing the sensitivity of the experiment to the level of 10@xmath21gev@xmath4 , probing the qcd axion model region for masses higher than 10@xmath22ev . in order to exceed the limit of 10@xmath21gev@xmath4 with a helioscope , specially developed magnets are necessary . taking advantage of the fact that the important parameter is the strength and not the homogeneity of the magnetic field , a stronger magnet with bigger aperture , could be designed and constructed especially for such an experiment . this is a more expensive and long term option , which would allow reaching the limits of the helioscope axion searches with an increased discovery potential . these different options as well as expected sensitivities were presented in detail in @xcite . the cast experiment has established the most stringent experimental limit on axion coupling constant over a wide range of masses , exceeding astrophysical constraints . the @xmath10 phase has allowed to enter in an unexplored region favoured by the theory axion models . from the absence of excess x - rays when the magnet was pointing to the sun , we set a preliminary upper limit on the axion - photon coupling of @xmath23 at 95% cl for @xmath24 ev , the exact result depending on the pressure setting . at present , with the @xmath12 run we are exploring deeper this region to reach sensitivities of @xmath171.2ev . the collaboration is looking into developping the new generation of helioscopes in order to reach sensitivities of the order of 10@xmath21gev@xmath4 leading to explore a large part of the qcd favoured model region including the otherwise non - accessible sub - kev range .	the status of the solar axion search with the cern axion solar telescope ( cast ) will be discussed . results from the first part of cast phase ii where the magnet bores were filled with @xmath0 gas at variable pressure in order to scan @xmath1 up to 0.4 ev will be presented . from the absence of excess x - rays when the magnet was pointing to the sun , we set a typical upper limit on the axion - photon coupling of @xmath210@xmath3gev@xmath4 at 95% cl for @xmath1@xmath5 ev , the exact result depending on the pressure setting . our search for axions with masses up to about 1.2 ev using @xmath6 as a buffer gas is , since last year , in progress in the second part of cast phase ii . expectations for sensibilities will be given . near future perspectives as well as more long term options for a new helioscope experiment will be evoked .
let @xmath1 be a smooth , compact riemannian manifold without boundary . let @xmath2 be a smooth anosov vector field and let us denote by @xmath3 the flow on @xmath4 generated by @xmath5 and let @xmath6 be a smooth potential function . then we can define the following differential operator @xmath7 it is a well established approach to study the dynamical properties of anosov flows by the discrete spectrum of the operator @xmath8 , the so called pollicott - ruelle resonances . the fact that for volume preserving flows and real valued potentials , the operator @xmath8 is an unbounded , essentially self adjoint operator on @xmath9 might suggest , that @xmath8 has good spectral properties on @xmath9 . however , due to the lack of ellipticity @xmath8 has mainly continuous spectrum which carries little information on the dynamics of the flow . a very important progress was thus to construct banach spaces @xcite or hilbert spaces @xcite for anosov flows in which the operator @xmath8 has discrete spectrum in a sufficiently large region ( see also @xcite for analogous results for anosov diffeomorphism ) . more precisely , it has been shown that there is a family of hilbert spaces @xmath10 parametrized by @xmath11 such that for any @xmath12 and for sufficiantly large @xmath13 the operator @xmath8 acting on @xmath10 has discrete spectrum in the region @xmath14 . this discrete spectrum is known to be intrinsic to the anosov flow together with the potential function and does not depend on the sufficiantly large parameter @xmath13 ( see section [ sec : micro_ruelle ] for a more precise statement ) . accordingly we call @xmath15 a _ pollicott - ruelle resonance _ if it is an eigenvalue of @xmath8 on @xmath10 for sufficiently large @xmath13 and we call @xmath16 the space of _ pollicott - ruelle resonant states_. as @xmath17 is not anymore a normal operator on @xmath10 , there might exist also exist finite dimensional jordan blocks and given a ruelle resonance @xmath18 we denote by @xmath19 the maximal size of a corresponding jordan block and we call @xmath20 the space of _ generalized pollicott - ruelle resonant states_. the interest in these pollicott - ruelle resonances and their resonant states arises from the fact that they govern the decay of correlations . in order to illustrate this in more detail , let us for the moment assume that the anosov flow is contact , which includes for example all geodesic flows on compact manifolds with negative sectional curvature . if we denote by @xmath21 the invariant contact volume form and take two smooth functions @xmath22 , then one is interested in the behavior of the correlation function @xmath23 in this case the following expansion of the correlation function has been established in ( * ? ? ? * corollary 1.2 ) ( see also ( * ? ? ? * corollary 5 ) ) @xmath24 here @xmath25 are the pollicott - ruelle resonances for a vanisihing potential ( @xmath26 ) which are lying in a strip @xmath27 . furthermore @xmath28 are the corresponding left and right generalized eigenstates and the sum over @xmath29 arises from the possible existence of jordan blocks of size @xmath30 . as @xmath31 all terms except the first one vanish exponentially in the limit of large times . this effect is interpreted as an exponential convergence towards equilibrium , known as exponential mixing and has first been established in the seminal works of dolgopyat @xcite and liverani @xcite . the way towards equilibrium is governed by the sub leading resonances and resonant states . while the resonances determine the possible decay modes , the resonant states determine the coefficients . if for example there would be a simple pollicott - ruelle resonance @xmath25 with a resonant state , that vanishes on an open set @xmath32 , then for all observables @xmath33 this resonance would not appear in the correlation expansion ( [ eq : cor_expansion ] ) . the main theorem of this article states , that this can not be the case . [ thm : support ] let @xmath3 be a topologically transitive @xmath0 anosov flow . let @xmath18 be a generalized pollicott - ruelle resonance and @xmath34 then @xmath35 . note that contrary to the correlation expansion ( [ eq : cor_expansion ] ) which we mentioned as a motivation , our result for the resonant states holds for a general @xmath0 anosov flow with no assumption on contact structures , smooth invariant measures , or even mixing properties . taking a suspension of a topological transitive anosov diffeomorphism , the theorem directly implies an analogous result for the resonant states of the anosov diffeomorphism . to our knowledge there has up to now little been known about the structure of pollicott - ruelle eigenstates . however in the proof of ( * ? ? * theorem 5.1 ) gouzel and liverani show an analogouse result for the peripheral spectrum of topologically mixing hyperbolic maps . in this case , there is however only one simple peripheral eigenvalue so the statement applies only to this one resonant state which corresponds to the equilibrium measure . furthermore for the particular case of a geodesic flows on manifolds of constant negative curvature , dyatlov , faure and guillarmou @xcite proved an explicit relation between pollicott - ruelle resonant states and laplace eigenstates . using this relation allows to transfer well established support properties of laplace eigenstates to support properties of pollicott - ruelle eigenstates . the article is organized as follows : in section [ sec : notation ] we introduce the basic definitions and recall some known facts about anosov flows and the microlocal approach to pollicott - ruelle resonances . section [ sec : support ] is devoted to the proof of the main theorem [ thm : support ] . _ acknowledgements : _ i would like to thank heiko gimperlein for enlightening discussions about the microlocal techniques in this proof . furthermore i am grateful to sebastin gouzel , stphane nonnenmacher , viviane baladi and gabriel rivire for the hints concerning the dynamical systems literature and in particular for the explications about the smoothness of the conditional measures ( c.f . theorem [ thm : anosov_fubini ] ) by the first two of them . last but not least , i thank colin guillarmou and joachim hilgert for stimulating discussions , from which this work emerged as well as for their detailed feed - back on an earlier version of the manuscript . this work has been supported by the grant dfg hi 412 12 - 1 . let @xmath1 be a smooth , compact riemannian manifold without boundary and let us denote the riemannian distance between two points @xmath36 by @xmath37 . a smooth vector field @xmath2 , respectively the flow @xmath3 on @xmath4 generated by @xmath5 , are called _ anosov _ if there is a direct sum decomposition of each tangent space @xmath38 which depends continuously on the base point @xmath39 and which is invariant under @xmath40 . furthermore the decomposition has to fulfill @xmath41 and there has to exist some fixed @xmath42 and @xmath43 with @xmath44 an anosov flow is said to be _ topological transitive _ if there exists a dense orbit . it is known , that the bundles @xmath45 and @xmath46 are uniquely integrable @xcite and given a point @xmath47 we denote by @xmath48 and @xmath49 the integral manifolds through the point @xmath39 which are smooth immersed submanifolds of @xmath4 . they are called _ weak stable _ and _ weak unstable _ manifolds through @xmath39 . also the bundles @xmath50 and @xmath51 are uniquely integrable . the corresponding integral manifolds will be denoted by @xmath52 and will be called _ strong stable _ and _ strong unstable _ manifolds . for further reference let us denote by @xmath53 the dimension of the manifolds @xmath54 and @xmath55 , respectively . let us recall the following well established result on the continuous dependence of the invariant manifolds on the base point . [ thm : cont_depend_inv_mflds ] @xmath56 is a family of @xmath57-dimensional immersed @xmath0-submanifolds that depends continuously on the base point w.r.t . the @xmath0-topology . more precisely , this means that for any @xmath39 there is a neighborhood @xmath58 as well as a continuous map @xmath59 such that for any @xmath60 @xmath61 is a @xmath0-diffeomorphism of the @xmath57 dimensional disk @xmath62 onto a neighborhood of @xmath63 in @xmath64 . the analogous result holds for @xmath65 . the stable and unstable foliations are known to be even hlder continuous in general , and under certain assumptions such as curvature pinching even regularity @xmath66 or better can be obtained ( c.f . * section 2.3 ) ) . we will , however , not explicitly need these refined regularity estimates in the sequel . given two points @xmath67 we denote by @xmath68 the metric distance on @xmath49 coming from the metric inherited from the riemannian metric on @xmath4 . in the same way we can define @xmath69 and @xmath70 . we can now define the following different open balls of radius @xmath71 . @xmath72 [ thm : product_neighbour ] there exists @xmath73 independent of @xmath47 such that for any @xmath47 and @xmath74 the following maps @xmath75 are unambiguously defined , injective and homeomorphisms onto their images . for a proof see @xcite ( see also @xcite for a more detailed statement ) . given a point @xmath47 and @xmath76 we define the open product neighborhoods of @xmath39 @xmath77 and @xmath78 we will call such product neighborhoods also rectangular neighborhoods . these product neighborhoods have the important property that the defining homeomorphisms imply foliations by local invariant manifolds . in the sequel we will only need the foliations of @xmath79-rectangles which are given by @xmath80 where @xmath81 are the local strong stable leaves and @xmath82 the local weak unstable leaves . note that from theorem [ thm : cont_depend_inv_mflds ] , as well the strong stable leaves @xmath83 as the weak unstable leaves @xmath84 are smooth submanifolds of the product neighborhood . still the foliations are no smooth foliations as the trivialization map @xmath85 is only continuous , which reflects the fact from theorem [ thm : cont_depend_inv_mflds ] that the local invariant manifolds @xmath83 ( @xmath84 respectively ) , depend only continuously on their basepoints @xmath86 ( @xmath87 respectively ) . as @xmath4 is a riemannian manifold , the leaves @xmath83 ( @xmath84 respectively ) inherit a riemannian metric and thus also a lebesgue measure which we denote by @xmath88 ( @xmath89 respectively ) . one way to introduce sinai - ruelle - bowen ( srb ) measures is to demand , that they are absolute continuous w.r.t . the weak unstable foliation . a measure @xmath90 on @xmath4 which is invariant under the anosov flow @xmath91 is called a _ srb - measure _ if for any product neighborhood @xmath92 and for all @xmath93 there are positive , measureable functions @xmath94 on @xmath84 as well as a measure @xmath95 on @xmath96 such that the restriction of @xmath90 to @xmath92 is given by @xmath97 more precisely this equality means , that for any @xmath98 @xmath99 there are other equivalent possibilities to define srb measure in terms of metric entropy and ergodic averages , see e.g. @xcite for an overview . for any anosov flow there exists a srb measure . if the anosov flow is topological transitive , the srb measure is unique and the anosov flow is ergodic w.r.t . the srb measure . for an original proof of this result in the context of axiom a diffeomorphism see @xcite and for generalizations to flows @xcite . let us collect some facts about the spectral theory of pollicott - ruelle resonances . suitable banach spaces , in which the differential operator @xmath8 , has discrete spectrum , have first been introduced by liverani @xcite . in this article we will use the microlocal approach to transfer operators , which has been introduced by faure , roy and sjstrand in a series of papers @xcite . the following result for anosov flows has originally been shown by faure and sjstrand ( * ? ? ? * theorem 1.4 ) . the reader might also be interested in ( * ? ? ? * proposition 3.2 ) where an alternative proof , using different microlocal techniques , is given . while these two publications only mention the case of zero potential a more general statement including arbitrary smooth potentials can be found in @xcite . [ thm : fredholm_property ] for any @xmath12 there is an @xmath11 and hilbert spaces @xmath10 and @xmath100 such that in the region @xmath14 the operator @xmath101 is fredholm of index 0 . here the hilbert space @xmath10 is an anisotropic sobolev space that fulfills the relations @xmath102 where @xmath103 denotes the ordinary sobolev space . the hilbert spaces @xmath100 can be defined by considering the operator @xmath8 acting on @xmath104 and we set @xmath105 which is a hilbert space with the norm @xmath106 since @xmath107 is invertible for @xmath108 large enough ( see e.g. ( * ? ? ? * lemma 3.3 ) or ( * ? ? ? * prop 3.1 ) ) , analytic fredholm theory implies that the resolvent @xmath109 has a meromorphic continuation to @xmath14 and the _ pollicott - ruelle resonances _ are defined to be the poles of the meromorphic continuation . given a pollicott - ruelle resonance @xmath18 the space @xmath110 is nonempty and its elements are called the corresponding _ pollicott - ruelle resonant states_. furthermore there is an integer @xmath111 determining the maximal size of a jordan block with spectral value @xmath18 , i.e. we have for all @xmath112 @xmath113 and we call the space @xmath20 the space of _ generalized pollicott - ruelle resonant states_. note that the fact that @xmath114 is densely contained in all @xmath10 together with the uniqueness of meromorphic continuation imply that neither the position of the resonance @xmath18 nor the spaces @xmath115 and @xmath116 depend on the parameter @xmath13 , so they are intrinsic objects of the anosov flow ( c.f . * theorem 1.5 ) ) . a central ingredient in the proof of theorem [ thm : fredholm_property ] and the construction of the anisotropic sobolev spaces is to study the symplectic lift of the flow action to @xmath117 . faure and sjstrand therefore introduce the decomposition @xmath118 which is defined such that @xmath119 note that if one identifies @xmath120 with @xmath121 via the riemannian metric this definition seems counterintuitive , as @xmath122 is identified with @xmath123 and vice versa . the reason for this choice in @xcite is that it is natural from the point of view of the symplectic lift of @xmath3 to @xmath117 ( c.f . * eq . ( 1.13 ) ) ) . we will not need further details on the construction of the anisotropic sobolev spaces and refer to the excellent expositions in @xcite . however , we will crucially use the following consequence of these constructions on the microlocal regularity of the pollicott - ruelle resonant states ( * ? ? ? * theorem 1.7 ) @xmath124 where @xmath125 denotes the space of distributions on @xmath4 whose wavefront sets are contained in some closed cone @xmath126 ( see ( * ? ? ? * lemma 8.2.1 ) for further details ) . in this section , we will prove theorem [ thm : support ] . let us start by giving a brief outline of the proof : as a first step , we will introduce suitable charts that are adapted to the foliation of @xmath4 into stable and unstable manifolds . then we study some particular distributions , denoted by @xmath127 , which are defined in these charts and which correspond to characteristic functions of tubes along the strong stable foliation . proposition [ prop : foruier_estimate ] then gives estimates on the decay of the fourier transformation of @xmath127 . as a consequence , one deduces that the wavefront set of the distributions @xmath127 are bounded away from @xmath128 , thus they can be paired ( or multiplied ) with pollicott - ruelle resonant states . using the fourier estimates in proposition [ prop : foruier_estimate ] we then show , that any pollicott - ruelle resonant state that does not have full support , has to be identically zero . therefore we only use two properties of the resonant states : the flow invariance of their support and their microlocal regularity . a major technical difficulty in the proof comes from the fact that the stable and unstable foliations for general anosov flows are only hlder regular . we call a finite open cover @xmath129 a cover of product neighborhoods if there are @xmath130 , @xmath131 such that @xmath132 ( c.f . theorem [ thm : product_neighbour ] ) . in order to simplify the notation we write @xmath133 , @xmath134 and denote the homeomorphism of theorem [ thm : product_neighbour ] by @xmath135 . recall that @xmath136 is a smooth submanifold . now let us construct a smooth atlas of @xmath4 by choosing @xmath0-charts @xmath137 for our open cover of product neighborhoods such that @xmath138 , scaledwidth=100.0% ] we can then define the open set @xmath139 and directly see that @xmath140 defines a homeomorphism between @xmath141 and @xmath142 . now for any borel set @xmath143 we define @xmath144 which is again a borel subset . note that for a point @xmath145 the sets @xmath146 are exactly the image of the local strong stable manifolds under the coordinate chart @xmath147 @xmath148 according to theorem [ thm : cont_depend_inv_mflds ] the @xmath149 are @xmath0-submanifolds depending continuously on the basepoint @xmath86 w.r.t . the @xmath0 topology . furthermore , the definition of @xmath150 allows us to associate to any borel set @xmath151 the distribution @xmath152 by setting for @xmath153 @xmath154:= \int_{v_i}\mathds{1}_{s_a}(z)\psi(z ) dz.\ ] ] given a chart @xmath147 we write @xmath155 and we define @xmath156 to be the minimal closed cones such that @xmath157 and @xmath158 note that for @xmath159 one has @xmath160 and that on the other side @xmath161 are complementary cones . thus from the continuity of the foliations and after a possible refinement of the charts we can assume that @xmath162 we now want to study the fourier transform of @xmath127 : [ prop : foruier_estimate ] let @xmath163 be an arbitrary cutoff function , then there is a constant @xmath164 such that for any borel set @xmath151 , @xmath165 uniformly in @xmath166 . here @xmath167 is the @xmath168-dimensional euclidean volume of @xmath169 . furthermore , fix a closed cone @xmath170 such that @xmath171 . then for all @xmath172 there is a @xmath173 such that for all borel sets @xmath151 @xmath174 uniformly for @xmath175 . in order to prove this theorem we use the definition of the fourier transform for compactly supported distributions and get @xmath176\| = \left\| \int_{v_i } \mathds{1}_{s_a}(z ) \chi(z)e^{-iz\xi}dz\right\|.\ ] ] in order to obtain the estimates for this integral let us recall the following fact which is a consequence of the absolute continuity of the strong stable foliation . [ thm : anosov_fubini ] for any @xmath177 defined as above and any @xmath145 there is a smooth function @xmath178 such that for any @xmath179 @xmath180 here @xmath181 are the euclidean measures on @xmath182 , @xmath183 is the restriction to the smooth submanifold @xmath184 and @xmath185 the induced measure on the submanifold @xmath184 . the functions @xmath186 are called _ conditional densities _ of the strong stable foliation . furthermore , their dependence on @xmath145 is continuous in the following sense : if @xmath187 is the set of normalized tangent vectors in @xmath188 , then for any @xmath189 the following @xmath29-norm @xmath190 is finite and for all @xmath189 the map @xmath191 is continuous . the existence of the conditional measures can be derived by a standard argument from the absolute continuity of the strong stable foliation ( see appendix [ sec : appendix ] respectively standard dynamical systems literature , e.g. ( * ? ? ? * section 6.2 ) ) . in order to obtain the smoothness of the conditional measures as well as the continuous dependence on the base point @xmath86 one additionally needs the smoothness of the jacobians of holonomy maps along the strong stable leaves . this statement can for example be found in ( * ? ? ? * appendix e ) . first we use the fubini formula for the strong stable foliation ( [ eq : anosov_fubini ] ) in order to estimate the integral in ( [ eq : ft_integral ] ) we obtain @xmath192 now the uniform bounds on @xmath193 directly imply ( [ eq : rho_ft_decay1 ] ) . in order to see ( [ eq : rho_ft_decay2 ] ) we use the boundedness of @xmath194 together with the following partial integration argument : for @xmath195 with @xmath196 and @xmath145 we define the tangent vector field @xmath197 here @xmath198 is the orthogonal projection onto @xmath199 which makes sense after having identified @xmath200 . from the definition of @xmath184 and @xmath201 we conclude that for any @xmath145 and @xmath188 we have have @xmath202 . thus there is @xmath203 such that @xmath204 uniformly in @xmath188 and @xmath145 . by continuity we find an open neighborhood @xmath205 of @xmath206 such that @xmath207 uniformly in @xmath208 , @xmath188 and @xmath145 . now we use the closedness of @xmath209 and a compactness argument to conclude that there are @xmath210 as well as a constant @xmath211 such that for any @xmath212 there is @xmath213 , such that @xmath214 uniformly in @xmath145 , @xmath188 . furthermore theorem [ thm : cont_depend_inv_mflds ] implies that @xmath215 . even if this function is not compactly supported , the fact , that the @xmath184 are precompact , and that these functions can always be smoothly extended to a slightly larger stable leaf implies that for all @xmath189 the norm @xmath216 is finite and depends continuously on @xmath145 . now consider @xmath217 performing @xmath218 times partial integration w.r.t . the differential operator @xmath219 we get @xmath220 where @xmath221 is independent of @xmath145 . note that for this independence we crucially use that @xmath222 as well as @xmath223 are continuous . recall that we can choose @xmath209 to be an arbitrary cone with @xmath224 . thus ( [ eq : rho_ft_decay2 ] ) implies that for any borel set @xmath151 we have @xmath225 , i.e. the wavefront set of @xmath127 is contained in @xmath201 . multiplication of @xmath127 with an arbitrary compactly supported distribution @xmath226 is therefore well defined and yields a compactly supported distribution @xmath227 . [ lem : rho_a_prod_estimate ] for each distribution @xmath226 there is a constant @xmath228 such that for all borel sets @xmath151 we have @xmath229 \right\| \leq c_v{\textup{vol}}_{n_{wu}}(a).\ ] ] as @xmath230 has compact support in @xmath231 let us choose a cutoff function @xmath163 with @xmath232 in a neighborhood of @xmath233 . then we calculate @xmath234 & = ( v\cdot(\chi\rho_a))[1]=\widehat{v\cdot(\chi\cdot \rho_a)}(0)=(\widehat v \star \widehat{\chi\cdot\rho_a})(0)\\ & = \int_{{\mathbb r}^n } \widehat v(\xi ) \widehat{\chi\cdot\rho_a}(\xi)d\xi \end{aligned}\ ] ] and note that the last integral converges as @xmath235 . now let us fix a closed cone @xmath170 fulfilling @xmath236 and @xmath237 . as @xmath238 we get for any @xmath239 a constant @xmath240 such that @xmath241 uniformly for @xmath242 . as any distribution is of finite order we in addition get the existence of an integer @xmath243 and a constant @xmath244 such that @xmath245 uniformly for all @xmath166 . splitting the integral over @xmath246 we thus obtain @xmath247\|\leq c_{n , v } \int_{{\mathbb r}^n\setminus \omega } \langle\xi\rangle^{-n}\left\|\widehat{\chi\cdot\rho_a}(\xi)\right\|d\xi + c_{v,2 } \int_{\omega } \langle\xi\rangle^{k_v}\left\|\widehat{\chi\cdot\rho_a}(\xi)\right\|d\xi\ ] ] for any given @xmath172 . then we can use ( [ eq : rho_ft_decay1 ] ) in the first integral and ( [ eq : rho_ft_decay2 ] ) with @xmath248 in the second integral which leads to the desired estimate ( [ eq : rho_a_prod_estimate ] ) . next we study the local structure of @xmath249 for a resonant state @xmath250 . therefore note that @xmath251 is a @xmath3-invariant closed subset . this is the only property that we use in the following lemma . [ lem : full_measure_tubes ] let @xmath252 be a closed subset invariant under the topological transitive anosov flow . if @xmath253 , then there exists a finite cover of product neighborhoods @xmath129 as well as for any @xmath254 an open subset @xmath255 such that @xmath256 where @xmath257 is the lebsegue measure on @xmath141 . as @xmath258 is nonempty and open , there exists @xmath259 and @xmath260 such that @xmath261 . let us introduce the open sets @xmath262 now take a cover of product neighborhoods @xmath129 with @xmath263 . we now want to consider the backwards iterates @xmath264 for @xmath265 . firstly , the flow invariance of @xmath266 assures , that @xmath267 . secondly the dynamical properties of the flow assures that the sets @xmath264 are stretched into the strong stable direction and thirdly , if @xmath90 denotes the srb measure , then ergodicity implies that @xmath268 . we will subsequently use these three facts in order to construct the sets @xmath269 . therefore we first consider for an arbitrary index @xmath270 and @xmath265 the connected components of @xmath271 . if a connected component can be written in the form @xmath272 with @xmath273 open , we call it a complete empty tube . we call them `` empty '' as these sets have no intersection with @xmath266 and we intend to build our set @xmath269 from the sets @xmath274 . note that for large @xmath275 the stretching and folding will imply that @xmath271 has many connected components . but as we started with a product neighborhood @xmath276 and as the strong stable foliation is preserved by the flow , there are maximally two connected components which are not complete empty tubes . they correspond to the two ends of the product neighborhood ( c.f . figure [ fig : tubes ] ) . we can however make them complete by considering the connected components @xmath277 . as @xmath278 , these connected components are also empty . furthermore the fact that @xmath279 together with the fact , that the product neighborhoods are stretched along the strong stable direction for negative times implies the following statement : any connected component of @xmath277 that contains a connected component of @xmath277 is a complete tube . thus we can define an open set @xmath280 such that the union of connected components of @xmath277 that contain a connected component of @xmath271 is given by @xmath281 with this definition we obviousely have @xmath282 we now can define the opens set @xmath283 and obtain @xmath284 as @xmath285 is a diffeomorphism , it only remains to prove that @xmath286 . this , however , is an immediate consequence of @xmath287 ( which uses the ergodicity w.r.t . the srb measure ) as well as the following lemma . [ lem : neighbour_prod_measure ] let @xmath90 be the srb measure , and @xmath288 with @xmath289 then @xmath290 . using the fact that the srb measure is absolutely continuous w.r.t . the weak unstable foliation we can write @xmath291 now , let us analyze the function @xmath292 . for @xmath293 we known from the definition of our charts , that @xmath294 is diffeomorphic to @xmath142 . as @xmath295 comes from a riemannian metric this induced measure is equivalent to lebesgue measure on @xmath142 and thus the assumption @xmath289 implies @xmath296 . as @xmath297 is a positive density this implies @xmath298 . now let us consider an arbitrary other point @xmath299 . we want to show , that also the sets @xmath300 are of positive measure w.r.t . this however is a direct consequence of the fact that @xmath302 has positive measure together with the fact , that the strong stable holonomy maps are absolutely continuous ( see e.g. ( * ? ? ? * section 8.6.2 ) ) . again the positivity of the density @xmath303 implies @xmath304 . @xmath305 is thus a strictly positive measurable function and consequently @xmath306 . we are now able to prove theorem [ thm : support ] : let @xmath18 be a pollicott - ruelle resonance , @xmath307 an associated generalized resonant state and suppose that @xmath308 . let @xmath309 be the unique integer , such that @xmath310 as the differential operator @xmath8 is local we have @xmath311 . furthermore because @xmath312 , the support @xmath313 is invariant by the ansosov flow so we can apply lemma [ lem : full_measure_tubes ] to @xmath314 . thus we obtain a cover @xmath129 of product neighborhoods as well as open subsets @xmath255 fullfilling @xmath315 after a possible further refinement we can also guarantee the existence of @xmath0-charts @xmath316 as well as closed cones @xmath317 fulfilling ( [ eq : un_stable_cones ] ) . now let us choose a smooth partition of unity @xmath318 and define @xmath319 . recall from ( [ eq : wavefront_cond ] ) that @xmath250 implies @xmath320 and by ( [ eq : un_stable_cones ] ) this implies that @xmath321 . now , using lemma [ lem : rho_a_prod_estimate ] , we obtain for an arbitrary borel set @xmath151 the estimate @xmath322\|\leq c_{u_i}{\textup{vol}}_{n_{wu}}(a)$ ] . we now want to use this estimate in order to conclude , that @xmath323 for all @xmath324 . now for any @xmath325 let us introduce the set @xmath326 as @xmath327 are open , also the family @xmath328 is a family of open sets , that increases monotonously and converges to @xmath269 for @xmath329 . considering the complement @xmath330 we obtain a monotonously decreasing family of closed sets converging to the nullset @xmath331 . furthermore , from ( [ eq : ni_suppu ] ) we see that the distribution @xmath332 defined in ( [ eq : rho_a ] ) is equal to @xmath333 on an open neighborhood of @xmath334 . taking an arbitrary @xmath153 and any @xmath325 we can use this fact to calculate @xmath335 = ( u_i\cdot \psi)[1 ] = ( ( u_i\cdot\psi)\cdot \rho_{n^c_{i,\varepsilon}})[1].\ ] ] but as the multiplication by a smooth function does nt increase the wavefront set we can apply lemma [ lem : rho_a_prod_estimate ] to @xmath336 and obtain for any @xmath337 @xmath338\| \leq c_{u_i,\psi } { \textup{vol}}_{n_{wu}}(n_{i,\varepsilon}^c).\ ] ] finally , as the constants @xmath339 are independent of @xmath337 , the right hand side converges to zero as @xmath329 , so we have shown @xmath340=0 $ ] . as @xmath153 was an arbitrary test function this implies @xmath341 for any @xmath270 and we conclude that @xmath342 which is a contradiction . this finishes the proof of theorem [ thm : support ] . in the proof of the support property of the pollicott - ruelle resonant states , the smoothness of the conditional measures ( theorem [ thm : anosov_fubini ] ) plays a central role . while the result is well known to experts in dynamical system theory it is difficult to find a precise statement or a proof in the existing literature . this is why i wanted to provide some more details in the arxiv - version of this article , on how to connect the smoothness of the conditional densities to the smoothness of holonomy maps , by standard dynamical systems arguments : let @xmath177 be the range of a coordinate chart @xmath147 as in theorem [ thm : anosov_fubini ] . as we only work in one fixed chart , we drop the index @xmath270 in order to simplify the notation and write @xmath343 . recall that we have already defined @xmath344 and analogousely we define @xmath345 . recall that the strong stable manifolds define a foliation of @xmath346 by the leaves @xmath184 which are parametrized by @xmath347 . note that @xmath348 so it can be identified with @xmath349 . now for any @xmath350 we define @xmath351 which defines a smooth foliation transversal to the foliation @xmath184 . recall that by the construction of the coordinate charts ( [ eq : chart_properties ] ) , @xmath352 coincides with a weak unstable leaf , but in general the other leaves can not coincide with the weak unstable leaves unless the particular case , that the weak unstable foliation is smooth . let us now reduce the set @xmath346 such that it has a product structure w.r.t . the foliations @xmath184 and @xmath353 , i.e. such that for all @xmath347 and @xmath350 one has @xmath354 defines a unique point contained in @xmath346 . note that this poses not problem for proving the theorem which claims the existences of the conditional measures in some product neighborhood of the strong stable and weak unstable foliation as such a set is always contained in the reduced set @xmath346 . now with this product structure of @xmath184 and @xmath353 we can define two holonomy maps @xmath355 and @xmath356 let us first discuss the second one , belonging to the smooth transversal foliation . according to theorem [ thm : cont_depend_inv_mflds ] , @xmath357 is a @xmath0 diffeomorphism and depends continuously on @xmath86 in the @xmath0-topology . note that the euclidean metric on @xmath346 induces metrics and thus measures on the submanifolds @xmath184 which we call @xmath358 and as @xmath359 is simply the @xmath87-axis @xmath360 is simply the euclidean measure @xmath361 on @xmath349 after identifying @xmath359 and @xmath349 . the fact that @xmath357 are diffeomorphisms gives us a jacobian function @xmath362 such that for a any function @xmath363 @xmath364 the theorem of change of variables gives us , the precise form of @xmath365 : for any @xmath366 the differential of the holonomy maps is a linear isomorphism @xmath367 choosing two orthonormal basis of the @xmath368-dimensional subspaces @xmath369 and @xmath369 ( w.r.t . to the euclidean metric on @xmath370 ) the differential can be expressed as an @xmath371 matrix and the jacobian is simply the absolute value of the determinant . now the fact , that @xmath372 is @xmath0 and depends continuously on @xmath86 translates to the fact that @xmath373 and depends continuously on @xmath86 w.r.t . the @xmath0-topology . let us now turn to the holonomy maps of the strong stable foliation . here the situation is infinitely more complicated , as for a general anosov flow the maps @xmath374 are only hlder continuous homeomorphisms . however it is known that they are absolutely continuous and that they possess a jacobian @xmath375 such that for any function @xmath376 @xmath377 the existence of such jacobians for holonomies in dynamical systems can be found in various textbooks ( see e.g. ( * ? ? ? * section 6.2 ) . it is however more difficult to find sharp statements on their regularity in @xmath86 and @xmath87 . the following sufficiently strong statement for our purpose can for example be found in ( * ? ? ? * appendix e ) let us now combine all the ingredients in order to calculate the conditional densities . given a function @xmath381 we calculate @xmath382dy\\ & = & \int_{v^s}\left [ \int_{t_0 } \psi\circ \mathcal h_y^s(x ) j_y^s(x ) d{\textup{leb}}_{t_0}(x)\right]dy\\ & = & \int_{t_0}\left[\int_{s_0 } \psi(\mathcal h(x , y ) ) j_y^s(x ) dm_{s_0}(y)\right]d{\textup{leb}}_{t_0}(x)\\ & = & \int_{t_0}\left[\int_{s_x } \psi(s ) \frac{j^s_{(\mathcal h^t_x)^{-1}(s)}(x)}{j^t_x((\mathcal h^t_x)^{-1}(s ) ) } dm_{s_x}(s)\right]d{\textup{leb}}_{t_0}(x).\end{aligned}\ ] ] we have thus shown that the conditional densities take the form @xmath383 collecting the statements on the regularity of the jacobians and holonomy maps we conclude , that for any @xmath86 they are smooth and that the map @xmath384 depends continuously on @xmath86 w.r.t . the @xmath0 topology . this statement assures , the uniform bounds on @xmath194 . in the case of @xmath0-foliations we could take coordinate charts @xmath147 such that @xmath385 . then ( [ eq : anosov_fubini ] ) would reduce to the ordinary fubini theorem . for a general foliation such generalizations of fubini theorems are a nontrivial task and there are counterexamples of foliations that are not absolutely continuous ( see e.g. ( * ? ? ? * section 8.6 ) ) .	we show that all generalized pollicott - ruelle resonant states of a topologically transitiv @xmath0 anosov flow with an arbitrary @xmath0 potential , have full support .
modern cosmology has already entered a precise , data - driven era . in 1998 , the elegant discovery that the universe is undergoing the phase of accelerated expansion @xcite , has motivated a great deal of studies concentrated on how best to parameterize the dark energy and measure its properties . an accompanying task comes naturally into being , i.e. , explore the corresponding behaviors at astrophysical scales when investigating the expansion history of the universe for different cosmological theories at cosmological scales . recently , because of the discovery of the accelerating universe , theorists have gradually paid more and more attention to the exotic spacetime configurations , especially , the renewed field wormholes . to be more concrete , in both cases ( accelerating universe and wormholes ) the null energy condition ( nec ) is violated and consequently all of other energy conditions . thus , an appealing overlap between two seemingly separated subjects occurs . so far , there is no doubt that together with black holes , white dwarfs , pulsars and quasars , etc . , wormholes constituting the most intriguing celestial bodies may provide a new window for physical discovery . in the literature , all the authors almost study the wormhole solutions only for a given dark energy model . however , with booming astronomical data , one can investigate better the wormhole spacetime configurations supported by cosmological observations through implementing highly precise constraints on a given dark energy model , in order to avoid choosing the values of the cosmological parameters arbitrarily . in our previous works @xcite , we have studied geometrical and holographical dark energy wormholes constrained by astrophysical observations for details , and verified that the exotic spacetime configurations wormholes can actually exist in our universe . it is worth noting that the obtained wormhole configurations are static and spherically symmetrical solutions existing at some certain low redshift . nonetheless , the previous results depend apparently on given dark energy models . hence , one shall ask whether the model - independent wormholes exist in the universe . in other words , whether can one investigate the existence of wormholes directly starting from cosmological observations ? to answer this question , one should determine the dark energy equation of state ( eos ) @xmath2 by using the model - independent methods , in order to satisfy the requirement @xmath3 which ensures the wormhole spacetime structure open . during the past few years , there were many model - independent methods to study the evolutional behaviors of dark energy eos @xmath2 in the literature , for instance , principal component analysis ( pca ) @xcite and gaussian processes ( gp ) @xcite , etc . in this work , we would like to use the gp method to reconstruct the dark energy eos @xmath2 and investigate the correspondingly model - independent wormhole solutions . the gp algorithm is a fully bayesian approach for smoothing data , and can preform a reconstruction of a function directly from data without assuming a parameterization of the function . as a result , one can determine any cosmological quantity from cosmic data , and the key requirement of the gp algorithm is just the covariance function which entirely depends on the cosmic data . in light of the special advantage , the gp method has been widely applied in exploring the expansion dynamics of the universe @xcite , the test of the base cosmological model @xcite , the cosmography @xcite , the distance duality relation @xcite , the determination of the interaction between dark energy and dark matter and cosmic curvature @xcite , dodging the cosmic curvature to probe the constancy of the speed of light @xcite , dodging the matter degeneracy to determine the dynamics of dark energy @xcite , the slowing down of cosmic acceleration @xcite , etc . it is noteworthy that the analysis in paper @xcite has indicated that the matrn ( @xmath4 ) covariance function is a better choice to carry out the reconstruction and it has been used in papers @xcite . the outline of the rest paper is as follows : in the next section , we make a brief review on gp method . in section iii , we reconstruct the dark energy eos @xmath2 in order to study the corresponding wormhole spacetime configurations . in section iv , we obtain two specific wormhole solutions including the cases of a constant redshift function and a linear shape function , and investigate the related physical properties . the discussions and conclusions are presented in the final section ( we use the units @xmath5 ) . in a spatially flat friedmann - robertson - walker ( frw ) universe , the luminosity distance @xmath6 can be expressed as @xmath7 where @xmath8 denotes the redshift , @xmath9 the present - day value of the hubble parameter and @xmath10 the dimensionless hubble parameter , respectively . subsequently , using the normalized comoving distance @xmath11 , the dark energy eos is written as @xmath12 } , \label{2}\ ] ] where the prime denotes the derivative with respect to ( w.r.t ) the redshift @xmath8 and @xmath13 is present - day value of the matter density ratio parameter . to study the model - independent wormhole configurations , firstly , one needs to reconstruct the dynamic dark energy eos @xmath14 . then , as described in our previous works @xcite , one shall concentrate on the parts in which the condition @xmath15 is satisfied and study the static and spherically symmetrical wormhole solutions at some certain redshift . the gp method can reconstruct a function directly from observational data without assuming a concrete parameterization for the function . here we utilize the package gapp ( gaussian processes in python ) @xcite to implement the reconstruction , which is firstly invented by seikel et al . for a pedagogical introduction to gp . usually , the gp is a generalization of a gaussian distribution which is the distribution of a random variable . in addition , the gp exhibits a distribution over functions . at each point @xmath16 , the reconstructed function @xmath17 is a gaussian distribution with a mean value and gaussian error . the key of the gp method is a covariance function @xmath18 which correlates the function @xmath17 at different reconstruction points . the covariance function @xmath18 depends entirely on two hyperparameters @xmath19 and @xmath20 , which characterize the coherent scale of the correlation in @xmath16-direction and typical change in the @xmath21-direction , respectively . in general , the choice is the squared exponential covariance function @xmath22 $ ] . however , the analysis in @xcite has verified that the matrn ( @xmath4 ) covariance function is a better choice to carry out the reconstruction . hence , we will adopt the matrn ( @xmath4 ) covariance function in the following analysis : @xmath23 . . \label{3}\ ] ] in our reconstruction , we use the union 2.1 data sets @xcite which consist of 580 sne ia data and cover the redshift range [ 0.015,1.4 ] . as noted in paper @xcite , we transform the distance modulus @xmath24 to @xmath25 in the following manner @xmath26 \label{4}\ ] ] with @xmath27 . furthermore , we set the initial conditions @xmath28 and @xmath29 in the reconstruction process . notice that the values of @xmath25 just depend on a combination of the absolute magnitude @xmath30 and @xmath9 . different from the previous literature @xcite , we adopt the latest 36 observational hubble parameter data ( ohd ) from the paper @xcite and the shift parameter @xmath31 from the recent planck s result @xcite as important supplements of sne ia data , since the gp algorithm in gapp does need the data of @xmath25 and @xmath32 at the same time . to be more precise , we conclude the relations between the above - mentioned data sets and the normalized comoving distance @xmath25 as follows @xmath33 @xmath34 @xmath35 where @xmath36 denotes the redshift of recombination and we use the values @xmath27 and @xmath37 @xcite . subsequently , we verify the correctness of our method by using the union 2.1 data sets to run the modified code . obviously , the result in fig . [ f1 ] is consistent with that in fig . 6 of the paper @xcite . and @xmath38 using the union 2.1 data sets . the shaded regions are reconstructions with @xmath39 and @xmath40 confidence level . the blue lines represents the underlying true model ( the mean value of the reconstructions ) . since this plot is aimed at verifying the correctness of our gp method , we take the labels @xmath41 and @xmath38 as distinctions with those @xmath42 and @xmath43 used in fig . [ f2 ] . ] and @xmath43 using a combination of the union 2.1 data sets , 36 ohd data points and planck s shift parameter . the 36 ohd data points are shown in the lower right panel . the blue and magenta lines represents the underlying true model ( the mean value of the reconstructions ) and the base cosmological model , respectively . ] using a combination of the union 2.1 data sets , 36 ohd data points and planck s shift parameter . the blue and magenta lines represents the underlying true model ( the mean value of the reconstructions ) and the base cosmological model , respectively . the left and right panels represent the longitudinal coordinate ranges [ -10,5 ] and [ -1.5,0 ] , respectively.,title="fig : " ] using a combination of the union 2.1 data sets , 36 ohd data points and planck s shift parameter . the blue and magenta lines represents the underlying true model ( the mean value of the reconstructions ) and the base cosmological model , respectively . the left and right panels represent the longitudinal coordinate ranges [ -10,5 ] and [ -1.5,0 ] , respectively.,title="fig : " ] in this section , we implement the reconstruction by using a combination of union 2.1 data sets , 36 ohd data points and planck s shift parameter , and the result is shown in fig . it is easy to find that the combined reconstructions @xmath44 and @xmath43 in fig . [ f2 ] give apparently a tighter constraint than those in fig . [ f1 ] , and that the base cosmological model is consistent with the gp reconstructions at @xmath45 level . more precisely , the ohd data provide a tighter constraint in the low redshift range , and the planck s shift parameter gives a substantially strict high - redshift constraint avoiding the divergence in the high - redshift range of fig . therefore , we can reconstruct the dark energy eos @xmath14 better in light of the stricter constraints on @xmath44 and @xmath43 . then , according to eq . ( [ 2 ] ) , we carry out the reconstruction in fig . [ f3 ] . to exhibit better the evolving behaviors of @xmath14 , we make two plots in the longitudinal coordinate ranges [ -10,5 ] and [ -1.5,0 ] , respectively . we find that we give a tighter constraint on @xmath14 by comparing fig . [ f3 ] with fig . 9 in paper @xcite . furthermore , we find that the base cosmological model is well compatible with our reconstruction at @xmath45 level , and that the dark energy eos @xmath14 may exhibit an oscillating behavior at @xmath46 level in the low redshift range [ 0 , 2 ] . concentrating on the parts @xmath3 in fig . [ f3 ] , this result implies that the evolving wormhole spacetime configurations supported by cosmological observations may actually exist in the universe . however , in the present paper , we just investigate the static and spherically symmetrical wormhole configurations at some certain redshift . consider a general metric for the wormhole models as follows @xmath47 where @xmath48 , @xmath49 , @xmath50 , @xmath51 and @xmath52 denote the radial coordinate , angular coordinates , the redshift function and shape function , respectively . usually speaking , there are four necessary ingredients to form a wormhole configuration @xcite : @xmath53 the nec is violated . @xmath53 satisfy the so - called flaring - out conditions , i.e. , @xmath54 , @xmath55 and @xmath56 when @xmath57 . @xmath53 @xmath58 shall be finite anywhere so that an horizon can be avoided . @xmath53 the asymptotically flatness condition shall be satisfied , i.e. , @xmath59 and @xmath60 when @xmath61 . in an orthonormal reference frame , using the einstein field equation , @xmath62 , we express the stress - energy scenario as : @xmath63 @xmath64 @xmath65 , \label{11}\ ] ] where @xmath66 , @xmath67 , @xmath68 and the prime denote the matter energy density , the radial pressure , the tangential pressure and the derivative w.r.t . @xmath48 , respectively . using the stress - energy conservation equation , @xmath69 , one can have @xmath70 which can be regarded as the relativistic euler equation or the hydrostatic equation of equilibrium for the matter threading a wormhole spacetime structure . subsequently , we would like to calculate the wormhole solutions by adopting the mean value @xmath71 , which lies in the @xmath46 range @xmath72 $ ] at redshift @xmath73 ( see fig . [ f3 ] ) . take into account a constant redshift function @xmath74 , substituting the dark energy eos @xmath75 and eq . ( [ 9 ] ) into eq . ( [ 10 ] ) , we obtain @xmath76 one can easily check that this shape function @xmath56 when @xmath57 , which satisfies the flaring - out conditions . subsequently , evaluating at the wormhole throat @xmath77 , we get @xmath78 . furthermore , taking the mean value @xmath71 from gp reconstruction , one can also get @xmath79 . because the constant redshift function @xmath80 is finite everywhere and @xmath59 when @xmath61 , this wormhole spacetime structure is both asymptotically flat and traversable . thus , the dimension of this wormhole configuration can be substantially large in principle . in what follows , we would like to analyze the the traversability of the wormhole configuration as our previous works @xcite . generally speaking , for a traveler in the spaceship who plans to journey successfully through the wormhole throat , there are three necessary ingredients [ 23 ] : @xmath53 the tidal acceleration shall not exceed 1 earth s gravitational acceleration @xmath81 . @xmath53 the acceleration felt by the travelers shall not exceed 1 earth s gravitational acceleration @xmath81 . @xmath53 the traverse time measured by the travelers and the observers who keep static at the space station shall satisfy the quantitative relations . more detailed descriptions can be found in paper @xcite . for simplicity , we only exhibit the key formula through some derivations in the following manner @xmath82 and the throat radius of the wormhole @xmath77 by assuming @xmath71 and @xmath83 m@xmath84s@xmath85 . in the medium panel , we exhibit the relation between the traversal velocity @xmath0 and the dark energy eos @xmath86 by assuming @xmath87 m and @xmath83 m@xmath84s@xmath85 . in the right panel , we exhibit the relation between the traversal velocity @xmath0 and the earth s gravitational acceleration @xmath81 by assuming @xmath87 m and @xmath71 . the shaded regions and the red ( dashed ) lines correspond to the allowable regions for the traversal velocity @xmath0 and the extreme cases the equality signs in inequality ( [ 14 ] ) . , title="fig : " ] and the throat radius of the wormhole @xmath77 by assuming @xmath71 and @xmath83 m@xmath84s@xmath85 . in the medium panel , we exhibit the relation between the traversal velocity @xmath0 and the dark energy eos @xmath86 by assuming @xmath87 m and @xmath83 m@xmath84s@xmath85 . in the right panel , we exhibit the relation between the traversal velocity @xmath0 and the earth s gravitational acceleration @xmath81 by assuming @xmath87 m and @xmath71 . the shaded regions and the red ( dashed ) lines correspond to the allowable regions for the traversal velocity @xmath0 and the extreme cases the equality signs in inequality ( [ 14 ] ) . , title="fig : " ] and the throat radius of the wormhole @xmath77 by assuming @xmath71 and @xmath83 m@xmath84s@xmath85 . in the medium panel , we exhibit the relation between the traversal velocity @xmath0 and the dark energy eos @xmath86 by assuming @xmath87 m and @xmath83 m@xmath84s@xmath85 . in the right panel , we exhibit the relation between the traversal velocity @xmath0 and the earth s gravitational acceleration @xmath81 by assuming @xmath87 m and @xmath71 . the shaded regions and the red ( dashed ) lines correspond to the allowable regions for the traversal velocity @xmath0 and the extreme cases the equality signs in inequality ( [ 14 ] ) . , title="fig : " ] where @xmath0 is the traversal velocity . subsequently , assuming the height of the traverser @xmath88 m , @xmath87 m , @xmath71 and @xmath83 m@xmath84s@xmath85 , we can have the traversal velocity @xmath89 m / s , which is approximately equivalent to 3 times the speed of sound . furthermore , we can also get the traverse time @xmath90 by setting the junction radius @xmath91 m. moreover , one can find that the value of the traversal velocity @xmath0 in inequality ( [ 14 ] ) depends on the throat radius of the wormhole @xmath77 , the dark energy eos @xmath86 from the gp reconstruction and the earth s gravitational acceleration @xmath81 . in addition , we are also of interest to investigate the dependence of the traversal velocity @xmath0 on the above - mentioned physical quantities . in the left panel of fig . [ f4 ] , it is easy to be seen that the traversal velocity @xmath0 is a linear function of the throat radius of the wormhole @xmath77 , meaning that with increasing @xmath77 , @xmath0 will also gradually increase . in the medium panel of fig . [ f4 ] , we find that the traversal velocity @xmath0 increase monotonically with gradually dark energy eos @xmath86 , which means that @xmath0 will increase with decreasing cosmic acceleration . notice that the dark energy eos @xmath86 will not approach @xmath92 since the right hand side of inequality ( [ 14 ] ) diverges . in the right panel of fig . [ f4 ] , assuming that the largest gravitational acceleration a human being can bear is 6.5 times earth s gravitational acceleration , we find that the traversal velocity @xmath0 still increase monotonically with increasing gravitational acceleration @xmath81 . take into account a specific shape function @xmath93 and utilize the dark energy eos @xmath75 , we have @xmath94 where @xmath95 denotes an arbitrary integration constant . we find that this solution is non - asymptotically flat since it diverges directly when @xmath61 . thus , this solution is non - traversable for an interstellar traveler . however , in theory , one can construct a traversable wormhole geometry through gluing an exterior flat geometry into the interior geometry at a junction radius @xmath96 . so the constant @xmath95 can be written as @xmath97 . subsequently , we will utilize the the method of `` volume integral quantifier '' ( viq ) @xcite to quantify the exotic matter constructing the traversable wormhole in the finite range @xmath98 . for simplicity , one can obtain the amounts of the exotic matter by calculating the definite integral @xmath99 , where @xmath100 is still the stress - energy tensor and @xmath101 any future directed null vector . in what follows , computing the quantity @xmath102dv$ ] , we obtain @xmath103'dr . \label{16}\ ] ] through some easy derivations , we get @xmath104 since the amounts of exotic matter @xmath1 is associated with the throat radius of the wormhole @xmath77 , the dark energy eos @xmath86 from the gp reconstruction and the junction radius @xmath96 , it is necessary to study the the dependence of the amounts of exotic matter @xmath1 on the above - mentioned physical quantities . in the left and right panels of fig . [ f5 ] , we find that @xmath1 increases with increasing @xmath77 and @xmath96 and approaches zero finally , since it is a linear function of @xmath77 and @xmath96 . in the medium panel of fig . [ f5 ] , however , we find that @xmath1 increases very slowly with increasing @xmath86 . this can be ascribed to the value of @xmath105 is very large . it is not difficult to verify that the amounts of exotic matter @xmath106 when @xmath107 and @xmath86 is fixed . for instance , if we still use the dark energy eos @xmath71 from gp reconstruction , eq . ( [ 17 ] ) can be rewritten as @xmath108 this result indicates that , in theory , one can construct a traversable wormhole with infinitesimal amounts of averaged - nec ( anec ) violating dark energy fluid in the present situation . it is worth noting that we have obtained the same result by using the model - independent gp method , which does not depend on any concretely cosmological model . and the throat radius of the wormhole @xmath77 by assuming @xmath71 and @xmath109 m. in the medium panel , we exhibit the relation between the amounts of exotic matter @xmath1 and the dark energy eos @xmath86 by assuming @xmath87 m and @xmath109 m. in the right panel , we exhibit the relation between the amounts of exotic matter @xmath1 and the junction radius @xmath96 by assuming @xmath87 m and @xmath71 . , title="fig : " ] and the throat radius of the wormhole @xmath77 by assuming @xmath71 and @xmath109 m. in the medium panel , we exhibit the relation between the amounts of exotic matter @xmath1 and the dark energy eos @xmath86 by assuming @xmath87 m and @xmath109 m. in the right panel , we exhibit the relation between the amounts of exotic matter @xmath1 and the junction radius @xmath96 by assuming @xmath87 m and @xmath71 . , title="fig : " ] and the throat radius of the wormhole @xmath77 by assuming @xmath71 and @xmath109 m. in the medium panel , we exhibit the relation between the amounts of exotic matter @xmath1 and the dark energy eos @xmath86 by assuming @xmath87 m and @xmath109 m. in the right panel , we exhibit the relation between the amounts of exotic matter @xmath1 and the junction radius @xmath96 by assuming @xmath87 m and @xmath71 . , title="fig : " ] the elegant discovery that the universe is in accelerating expansion has motivated theorists to pay afresh attention to the old field wormholes , since in both fields an surprising overlap occurs , i.e. , the violation of the nec . however , in the literature , all the authors almost study the wormhole spacetime configurations for a given cosmological model . therefore , in light of rapidly increasing cosmic data , we are aiming at exploring the model - independent dark energy wormholes by using the gp method . in this work , first of all , we make a brief review on the gp method and demonstrate the correctness of our reconstruction method . subsequently , using a combination of union 2.1 580 sne ia data , the latest 36 ohd data points and recent planck s shift parameter , we find that our gp method can give a tighter constraint on the normalized comoving distance @xmath110 , its derivatives @xmath111 and the dark energy eos @xmath14 than the previous literature @xcite . in the mean while , we find that the base cosmological model is well compatible with our reconstruction at @xmath45 level . in what follows , two specific traversable wormhole solutions are obtained in the framework of a perfect fluid , i.e. , the cases of a constant redshift function and a linear shape function . by choosing the mean value @xmath71 , which lies in the @xmath46 range @xmath72 $ ] at redshift @xmath73 ( see fig . [ f3 ] ) , we analyze the traversabilities of the former case and investigate the dependence of the traversal velocity @xmath0 on the throat radius of the wormhole @xmath77 , the dark energy eos @xmath86 from the gp reconstruction and the gravitational acceleration @xmath81 . furthermore , we also calculate the amounts of exotic matter of the latter case and study the the dependence of the amounts of exotic matter @xmath1 on the throat radius of the wormhole @xmath77 , the dark energy eos @xmath86 from the gp reconstruction and the junction radius @xmath96 . we find that , with decreasing cosmic acceleration , the traversal velocity @xmath0 of the former case increases and the amounts of exotic matter @xmath1 of the latter case decreases ( see figs . [ f4]-[f5 ] ) . the dark energy eos @xmath14 from our gp reconstruction may exhibit an oscillating behavior at @xmath46 level in the low redshift range [ 0,2 ] , which indicates that the evolving wormhole spacetime configurations supported by cosmological observations may actually exist in the universe . furthermore , the evolving wormhole spacetime structure originated from cosmic acceleration may open at a relatively high redshift and close at a low redshift based on some unknown mechanism ( e.g. , quantum fluctuations ) . according to the papers @xcite , the obtained wormhole configurations from our gp construction can be stable by choosing appropriately the values of the throat radius and junction radius . however , the interesting stability analysis is beyond the scopic of the present work . we expect more and more high quality cosmic data can help the human beings to explore the nature of dark energy and dark matter . the author xin - he meng warmly thanks professors bharat ratra and s. d. odintsov for helpful feedbacks on astrophysics and cosmology . the author deng wang thanks qi - xiang zou for beneficial discussions and programming . the authors acknowledge partial support from the national science foundation of china .	we study the model - independent traversable wormholes from cosmological reconstruction based on gaussian processes ( gp ) . using a combination of union 2.1 sne ia data , the latest observational hubble parameter data and recent planck s shift parameter , we find that our gp method can give a tighter constraint on the normalized comoving distance , its derivatives and the dark energy equation of state than the previous work @xcite . subsequently , two specific traversable wormhole solutions are obtained , i.e. , the cases of a constant redshift function and a linear shape function . we find that , with decreasing cosmic acceleration , the traversal velocity @xmath0 of the former case increases and the amounts of exotic matter @xmath1 of the latter case decreases .
since the discovery of the _ p - si _ photoluminiscence ( pl ) at room temperature realized by l.t . canham in 1990 @xcite , a lot of studies have been carried out to understand the origin of this property . the possibility that _ p - si _ could be applied to photonic and optoelectronic devices , motivated the study of the dependency of its pl to the temperature , chemical attack variables and molecules on its surface . despite the big amount of scientific reports about the _ p - si _ pl , until now , we have not a model that reproduce the characteristics of all experimental pl spectra reported in the bibliography . + during the research of the origins of the _ p - si _ pl , some models about the nature of the pl were proposed ; cullis _ _ et al.__@xcite say that all these models could be reorganized in 6 mayor groups . we only review the fundamentals of the confinement quantum model ( qc ) , and of the hydrogen - groups model ( hs ) because we will use them to interpret our results . + the qc model proposed by canham in 1990 was the first model created with the aim to explain the efficient pl of the _ p - si _ at room temperature . in this model , because of the effects of the quantum confinement , the energy gap of the silicon nanostructures increase , and their photoluminescence appear in the visible range of the electromagnetic spectrum . in accord to the qc model , if the size of the silicon nanostructures ( where the recombination process take place ) decrease , then the energy of the photons created increase . the qc model is the most successful theory about the nature of the _ p - si _ photoluminescence , because it explain various experimental observations , with an exception : the pl spectra of oxidized _ p - si _ structures . + in the other hand , the hs model consider that the hydrogen atoms on the _ p - si _ surface ( which satisfy the dangling bonds ) , could be involved in the origin of pl of this material . the hs model was proposed after observing that if the hydrogen atoms are removed from the _ p - si _ surface , by a thermic procedure , then the intensity of the pl decrease dramatically ( but still exist because of the qc model argument ) . since the _ p - si _ with a rusty surface present an intense pl , we understand that , the porous silicon with an hydrogen passivated surface ( _ p - si : h _ ) is just one of the luminescent forms of the _ p - si _ ( cullis _ et al . _ ) . + photoluminescence is a difficult phenomenon to simulate , since the dynamics of the excitation and decay of the electrons require precise calculations of the excited states and decaying mechanisms . also , metastable states may take part in the process with short , but relevant , mean free lives that require time - dependent quantum mechanical calculations @xcite , @xcite . instead , we have decided to initially investigate the photoemission spectrum as a crude approximation to the photoluminescence phenomenon . + bearing in mind the qc and hs models , we propose a simple photoemission model in which the excited electrons of the conduction band , decay successively into the unoccupied electronic states of the valence band ( see figure [ modeca ] ) . the emission spectra that we obtain for the _ p - si _ and _ p - si : h _ structures are comparable to the experimental spectra reported by cullis _ et al_. + we are aware that our simulated _ p - si _ structures do not have experimental counterpart , because the number of silicon atoms involved in each simulation is finite due to the computational cost of the ab initio calculations . also , in our simulations is not possible to reproduce all the chemical molecules that generate after the chemical attack of the _ p - si _ surface , because depending of the atmosphere where the attack was realized , molecules or atoms besides hydrogen atoms , could satisfy the dangling bonds of the _ p - si _ surface modifying the form of its pl spectrum . + however , until now the doubts about the origins of the _ p - si _ pl remains , so , our work contributes to calculate the photoemission spectra of _ p - si _ and _ p - si : h _ structures with ab initio molecular dynamics methods . in the following section , we describe the construction method used to generate the _ p - si _ and _ p - si : h _ structures , and the way we obtain their photoemission spectra . the _ p - si _ structures were created using the _ cerius2 _ interface @xcite . we reproduced the silicon cell diamond - like which contains 8 silicon atoms 2 times along each of the coordinate axes to obtain a 64 silicon - atom supercell , 3 times to increase the number of atoms to 216 , 4 times for a supercell with 512 atoms and finally , 5 times along each direction to obtain 1000 atoms in the silicon supercell . + once the supercells were created , we eliminated 50% of their silicon atoms carving a central pore parallel to the z - axis and with a regular cross section . after eliminating 50% of the silicon atoms contained in each supercell , we obtained 4 _ p - si _ structures with 32 , 108 , 256 and 500 silicon atoms . the pore of the _ p - si _ structure with 32 silicon atoms has an irregular transversal section because the great number of atoms removed from the supercell . in the other hand , the pore of the supercell with 500 silicon atoms shows a regular cross section because even if we remove 50% of their atoms , its silicon backbone is bigger than the one for the 32 silicon atoms model . + the 4 structures of _ p - si _ were subjected to a geometry optimization process with the objective of finding their equilibrium configurations ( see figures [ psi32 ] to [ psi500 ] ) . we perform the geometry optimization process with _ fast structure _ a code included in the _ cerius2 _ interface based in the lin and harris approach @xcite , @xcite , and the vwn @xcite functional of correlation energy . + in table [ uno ] we indicate the computational parameters used for the geometry optimizations of the 4 _ p - si _ structures . the integration grid refers to the space were the atomic functions , atomic densities , coulomb potential , exchange and correlation potentials are represented . to calculate the hamiltonian matrix elements it is necessary to put each atom in the center of the integration grid . a coarse integration grid leads to a poor representation of the atomic functions , whereas a fine grid is the best approximation but its computational cost is elevated . the minimal bases indicate that we use the valence atomic orbital to construct the wave function , and the minimal density indicates that the electronic density should be obtained just using the valence atomic orbital as well . finally , the cut - off parameter corresponds to the distance at which the wave function vanish . + at the end of the geometry optimizations of our _ p - si _ structures , the interface brings forth their radial distribution functions ( figures [ fdr32 ] to [ fdr500 ] and [ secuencia ] ) and densities of states . we used part of the values of the densities of states of the _ p - si _ models as the input information for our emission code ( edoxedo.f ) . we consider that the electronic states of the conduction density are unoccupied and those of the valence density are occupied . following the decay order proposed by our code ( figure [ modeca ] ) , the excited electrons of the conduction states decay into the valence density states emitting a photon , whose energy is equal to the energetic difference between the two electronic states involved . the energies of the photons are computed as an histogram and then , we adjust a gaussian curve to the histogram to compare it with the _ p - si _ pl experimental spectrum reported by cullis _ ( cullis _ et al . _ ) . to calculate the emission of all our structures , we only consider the electronic states that belong to a 2 ev interval in the valence and conduction densities of states . we decide to work out with a 2 ev interval states because experimentally , the majority of the pl excitation sources have energies of this order of magnitude . + the _ p - si : h _ models were constructed taking the _ p - si _ optimized structures as the initial atomic arrangement . we saturated with hydrogen the _ p - si _ optimized structures . after the saturation with hydrogen , we applied a geometry optimization to the _ p - si : h _ structures using the parameters of table [ uno ] . in figures [ psi32h42 ] to [ psi500h196 ] we present the 4 optimized structures of _ p - si : h_. we obtain the valence and conduction states from the densities of states of the _ p - si : h _ models in the same way than for the _ p - si _ structures . again , to obtain the emission spectra of the _ p - si : h _ models we just consider the electronic states , in the valence and conduction densities , that are in an 2 ev interval . the order of decay of the excited electronic states into the unoccupied valence states is shown in figure [ modeca ] . the distribution of the photon energies is obtained after a gaussian fit . + in the results section we exhibit the photoemission spectra of the _ p - si _ and _ p - si : h _ models . we compare our photoemission spectra with the pl spectrum reported by cullis _ et al._. .[uno ] . computational parameters used for the geometry optimization process implemented by the _ fast structure _ code for the _ p - si _ and _ p - si : h _ models . [ cols="^,^ " , ] . interatomic distance variation of the _ p - si _ optimized models , as a function of the size of their silicon backbones.,scaledwidth=90.0%,height=7 ] . order ( considered in the edoxedo.f code . ) , of the excited electrons decay from the conduction band to the valence band.,scaledwidth=70.0%,height=7 ] as we can see from figures [ fdr32 ] to [ fdr500 ] , the radial distribution functions of the _ p - si _ structures become similar to the radial distribution function of a crystalline silicon structure , until the number of silicon atoms reach 500 ( figure [ fdr500 ] ) . the variation of the interatomic distance of the _ p - si _ structures as a function of their silicon backbones size is shown in figure [ secuencia ] . the backbone of the _ p - si _ structure with 500 silicon atoms is thicker than the one of the model with just 32 silicon atoms , therefore , the thinner silicon backbone is really deformed after a geometry optimization process because of the presence of the porous in the simulation cell , then the most probable distance between silicon neighbors increase to 2.55 . in the other hand , the backbone of the _ p - si _ structure with 500 silicon atoms , is thick enough to remain as a crystalline silicon structure despite the central porous , so then the distance between first neighbors is 2.35 , as for a crystalline silicon structure . + figures [ ajus32 ] , [ ajus108 ] , [ ajus256 ] and [ ajus500 ] show the gaussian curves fitted to the histogram of each _ p - si _ model obtained by our edoxedo.f code . in figure [ plpsi ] we present the emission spectra ( just the gaussian curves ) of the 4 _ p - si _ optimized structures ; for these 4 structures their photoemission energy range from 1.2 ev to 2.8 ev , its mean maximum photoemission energy is 1.94 ev , and as we can observe , the intensity of the emission increase if the size of the backbone increase , but its maximum energy decrease in accord to the results of cullis _ et al._. + in figure [ cullispl ] we compare the _ p - si _ pl spectrum reported by cullis _ et al . _ and the emission spectrum of our _ p - si _ model with 500 silicon atoms ; from this figure we can appreciate that our spectrum is larger than the cullis _ et al . _ but both have a similar form . also , the maximum photoemission energy of our spectrum is 0.34 ev to the blue of the electromagnetic spectrum considering the va in figure [ plpsihmod ] we superpose the emission spectra of the 4 models of _ p - si : h _ calculated by the edoxedo.f code . we did not find experimental photoemission spectra for the _ p - si : h _ to compare to our results . emission spectra of the four _ p - si : h _ models calculated with de edoxedo.f code.,scaledwidth=49.0%,height=5 ] we observed that the form of the emission spectra , obtained after a gaussian fitting of the histograms calculated by the edoxedo.f code , is similar to the form of the cullis _ the maximum width of our spectra is 1.6 ev , their minimal energy is 1.2 ev and the maximal one is 2.8 ev . the mean emission energy calculated was 1.94 ev , this is , 0.34 ev toward the blue of the electromagnetic spectrum , with respect the cullis _ maximum energy . + is clear from the figure [ plpsi ] that decreasing the number of silicon atoms in the simulation cell , moves toward bigger energies the maximal energy of the emission spectra . we think that this fact is due to the size of the quantum confinement region in the silicon backbone ; for example , in the 32 _ p - si _ structure , the silicon backbone is smaller than the one of the 500 silicon atoms model and , in accord with the quantum confinement model , his emission energy should be bigger than the one of the _ p - si _ 500 atoms model . + from figure [ plpsi ] we conclude that decreasing the number of silicon atoms in the _ p - si _ supercells results in a smaller emission intensity . we think that because the _ p - si _ 500 silicon atoms model has a crystalline structure ( [ fdr500 ] ) compared with the 32 atoms structure ( [ fdr32 ] ) , the radiative recombination of the charges dominates in the crystalline structures of the _ p - si _ and is not significant in the amorphous ones . + we observe from figure [ cullispl ] that the emission spectra of our _ p - si _ 500 silicon atoms model is toward the right of the electromagnetic spectrum , with respect to the pl maximum reported by cullis _ et al._. we consider that the size of the supercell is the direct cause and probably , if we could be able to increase the number of atoms , to thousand of they to construct the _ p - si _ models , theirs emission maximus will be near the experimental ones reported . + emilye rosas landa loustau acknowledges the financial support of conacyt during his phd studies ; she also wants to thank dr . lorenzo pavesi , and dr . juan carlos alonso huitrn because of their fruitful discussions .	we created 4 _ p - si _ models and 4 _ p - si : h _ models all with 50% porosity . the models contain 32 , 108 , 256 and 500 silicon atoms with a pore parallel to one of the simulational cell axes and a regular cross - section . we obtained the densities of states of our models by means of ab initio computational methods . we wrote a code to simulate the emission spectra of our structures considering particular excitations an decay conditions . after comparing the simulated spectra with the experimental results , we observe that the position of the maximum of the emission spectra might be related with the size of the silicon backbone for the _ p - si _ models as the quantum confinement models say and with the hydrogen concentration for the _ p - si : h _ structures . we conclude that the quantum confinement model can be used to explain the emission of the _ p - si _ structures but , in the case of the _ p - si : h _ models it is necessary to consider others theories .
a number of alternative theories of gravity have been proposed in the last decade in relation with the cosmic speedup problem . among them , @xmath0 theories have attracted special attention . these theories have been mainly studied in the metric@xcite and in the palatini@xcite formalisms . to be in agreement with local tests of gravity , the former case generically requires that the non - linear curvature terms represent a short - range , massive scalar degree of freedom . the latter introduces non - perturbative effects that severely constrain the kind of non - linear terms , ruling out infrared corrections . here we present a hybrid combination@xcite of both approaches which exhibits interesting new features that allow to avoid the typical shortcomings of the pure metric and palatini @xmath0 theories . recent applications are briefly summarized . the hybrid metric - palatini action @xmath1 + s_m \ , \ ] ] where @xmath2 is the matter action , @xmath3 , @xmath4 is the einstein - hilbert term , @xmath5 is the palatini curvature , and @xmath6 is defined in terms of an independent connection @xmath7 as @xmath8 , admits the following scalar - tensor representation @xmath9+s_m .\ ] ] in the weak - field limit and far from the sources , the scalar field behaves as @xmath10 , whereas the metric perturbations @xmath11 take the form @xmath12 where @xmath13 , @xmath14,\qquad { \rm and } \qquad \gamma \equiv \frac{1+\left(\phi_0/3\right)e^{-m_\varphi r}}{1-\left(\phi_0/3\right)e^{- m_\varphi r}}\,.\ ] ] as is clear from these expressions , the coupling of the scalar field to the local system depends on the amplitude of the background value @xmath15 . if @xmath15 is small , then @xmath16 and @xmath17 regardless of the value of the effective mass @xmath18 . this contrasts with the result obtained in the metric version of @xmath0 theories @xcite and allows to pass the solar system tests even if the scalar field is very light as long as @xmath15 is sufficiently small . a light scalar field has a long interaction range . this implies that our theory can modify the long - range gravitational interaction without having any relevant impact on the dynamics of local systems . thus , the hybrid metric - palatini theory opens up new possibilities to approach the problems of both dark energy and dark matter . the cosmological applications of this theory have been investigated recently@xcite by the authors . criteria to obtain cosmic acceleration were analyzed , and the field equations were formulated as a dynamical system . several classes of cosmological solutions describing both accelerating and decelerating universes were explicitly obtained . furthermore , the cosmological perturbation equations were derived and applied to uncover the signatures of these models in the large - scale structure . + more recently , we have analysed the generalized virial theorem@xcite . taking into account the relativistic collisionless boltzmann equation , we showed that the new geometric terms in the field equations provide an effective contribution to the gravitational potential energy . indeed , it was shown that the total virial mass is proportional to the effective mass associated with the new terms generated by the effective scalar field , and the baryonic mass . this shows that the mass discrepancy in clusters of galaxies might be due to the effects of modified dynamics . we found that the mass associated to the scalar field and its effects extend beyond the virial radius of the clusters of galaxies . in the context of the galaxy cluster velocity dispersion profiles predicted by the our model , the generalized virial theorem can be an efficient tool in observationally testing the viability of this class of generalized gravity models . thus , hybrid metric - palatini gravity provides an effective alternative to the dark matter paradigm of present day cosmology and astrophysics . + from a purely theoretical perspective , the general conditions for wormhole solutions according to the null energy condition at the throat in the hybrid @xmath0 theory have also been considered@xcite . several wormhole type solutions were also obtained and analyzed . in the first solution , the redshift function and the scalar field were specified , and the potential was chosen so that the modified klein - gordon equation can be simplified . in the second example , by adequately specifying the metric functions and choosing the scalar field , we found an asymptotically flat spacetime . in a series of works we have shown that a combination of metric and palatini curvature terms yield a purely geometric theory of gravity in which a light scalar degree of freedom is able to modify the large scale structure without damaging the well - tested dynamics of local systems . the extra terms appearing in the gravitational field equations can describe the early accelerated dynamics of the universe and explain the matter deficit in galactic clusters . all the relevant astrophysical quantities , including the `` dark mass '' associated to the equivalent scalar - tensor description , can be expressed in terms of the physical parameters of the model the scalar field , its potential , as well as the coupling parameters . therefore , these results open the possibility of directly testing the hybrid modified @xmath0 type gravitational models by using direct astronomical and astrophysical observations . future work will focus on the dynamics of massive test particles at galactic and extra - galactic scales , and in the formal mathematical aspects of this theory . sc is supported by infn ( iniziative specifiche na12 and og51 ) . tsk is supported by the research council of norway . fsnl acknowledges financial support of the fundao para a cincia e tecnologia through the grants cern / fp/123615/2011 and cern / fp/123618/2011 . gjo is supported by the spanish grant fis2011 - 29813-c02 - 02 and the jae - doc program of the spanish research council ( csic ) . a. de felice and s. tsujikawa . living rev . * 13 * , 3 ( 2010 ) ; t. p. sotiriou and v. faraoni , rev . phys . * 82 * , 451 ( 2010 ) ; s. nojiri and s. d. odintsov , phys . rept . * 505 * , 59 ( 2011 ) ; f. s. n. lobo , arxiv:0807.1640 [ gr - qc ] . s.capozziello , t.harko , t.s.koivisto , f.s.n.lobo , g.j.olmo , arxiv:1209.2895 [ gr - qc ] . s. capozziello , t. harko , t. s. koivisto , f. s. n. lobo and g. j. olmo , arxiv:1212.5817 [ physics.gen-ph ] . s. capozziello , t. harko , t. s. koivisto , f. s. n. lobo and g. j. olmo , phys . d * 86 * , 127504 ( 2012 ) .	we present an extension of general relativity in which an @xmath0 term la palatini is added to the usual metric einstein - hilbert lagrangian . expressing the theory in a dynamically equivalent scalar - tensor form , we show that it can pass the solar system observational tests even if the scalar field is very light or massless . applications to cosmology and astrophysics , and some exact solutions are discussed .
in the recent years the largely forgotten idea that momentum space may have a nontrivial geometric structure , known under the name of born reciprocity @xcite , has been revived in many different guises in the context of quantum gravity . it was noticed in @xcite that there is a one - to - one correspondence between space - time noncommutativity , expected to be one of the features of quantum gravity , and nontrivial geometric structures in momentum space . this general observation is supported by explicit calculations done in the context of gravity in 2 + 1 dimensions @xcite , @xcite . few years later it was realized that many nontrivial features of doubly special relativity class of theories @xcite , @xcite , @xcite can be conveniently described in terms of the geometry of de sitter momentum space @xcite . relative locality @xcite , @xcite , @xcite is a theoretical framework that has its roots in born reciprocity . in this framework the momentum space is brought to foreground . it is first observed that most , if not all , physical measurements correspond , in fact , to momentum space data . second it is noticed that the emergence of a nontrivial geometry in momentum space requires , as a prerequisite , the presence of a mass scale . such scale must be provided by a fundamental theory , and it was assumed that there exists a regime of quantum gravity , in which the length scale , the planck length , is negligibly small , while the mass scale , the planck mass , remains finite . in the couple of years that passed since relative locality was first proposed the bulk of research investigated systems defined on flat minkowski spacetime . the question arises however if curved momentum space could coexists with a nontrivial geometry of spacetime . this possibility is particularly intriguing from the phenomenological perspectives of dsr and relative locality , since many of the opportunities that have been proposed in the recent years rely on tests of ( planck - scale ) deformation of kinematics of particles coming from cosmological distances @xcite,@xcite . the incredibly small size of such effects ( @xmath0 ) could be within the reach of present observations thanks to the huge amplification provided by the cosmological distances . in such context it is clear that the effects of spacetime curvature can not be ignored . some preliminary results on the interplay between spacetime expansion and relativity of locality have been presented in the paper @xcite , for the case of a de sitter - like spacetime expansion . recently two of the present authors proposed the action of a particle moving in curved spacetime , whose geometry is given by the tetrad @xmath1 , with curved momentum space , provided by the tetrad @xmath2 while the framework presented in @xcite reproduces the correct action in both the complementary limits of flat spacetime / curved momentum space and curved spacetime / flat momentum space , the theory described by the action proposed in @xcite is not manifestly invariant under general coordinate transformations ( at least in their classical form ) . this , by itself , may not be very problematic , because the scale governing the effects related to this loss of invariance would be , in this theory , presumably of order of the planck scale , and therefore they may not lead to relevant phenomenological consequences . however , from the theoretical perspective , one may find annoying the fact that the violation of the general coordinate invariance would lead to the presence of a preferred spacetime coordinate system , in which the particle action has the form proposed in @xcite . the problem would be then to find out what this coordinate system is : is it indeed the system of cartesian coordinates in minkowski space , in the case of flat spacetime , as implicitly assumed in the construction of relative locality ? and , more importantly , how to find such system , and the form of the particle action , in the case of an arbitrary spacetime ? one possibility to avoid this conflict could be to look for a generalized class of coordinate transformations under which the action presented in @xcite is still invariant . this would lead inevitably to coordinate transformations mixing spacetime and momentum space . taking into account that the existence of such kind of transformations is not guaranteed , it may also be conceptually compelling to explore the implications of having a theory of both spacetime and momentum space curved , in which invariance under general coordinate transformation is lost at the planck scale . while we postpone these analyses to future studies , we feel however that the option of a coordinate invariant theory is still the most desirable road to pursue in the search of generalizing relative locality to curved spacetime . thus , in this paper we present a novel formulation of the action for particles in both curved spacetime and momentum space , which is manifestly invariant under general coordinate invariance . we also show that in this theory the particles are described to move along worldlines coinciding with the standard spacetime geodesics . in the next section we show how one can construct such an action for free particles . in section 3 we discuss symmetries of so defined theory . in section 4 we show how , starting from this action , one can also introduce particle interactions in the spirit of relative locality . the final section is devoted to discussion . when this work was being completed we learned about an interesting complementary results presented in @xcite . the action of a free relativistic particle with curved momentum space has the form reciprocal , in a sense , to the one of the standard free relativistic particle moving in curved spacetime , with flat momentum space . the lagrangian of the latter reads @xmath3 where @xmath4 is the position of the particle at time @xmath5 , @xmath6 is the particle momentum , and @xmath7 is the tetrad , characterizing the geometry of spacetime @xmath8 finally @xmath9 is the lagrange multiplier enforcing the mass - shell constraint @xmath10 . to write the lagrangian ( [ 1 ] ) we use two kind of indices : the curved spacetime index @xmath11 and the index @xmath12 related to the orthonormal coordinate system in the ambient minkowski space , to which the tetrad @xmath7 maps . it can be checked by direct calculation that the euler - lagrange equations following from ( [ 1 ] ) reduce , after solving for @xmath13 , to the standard geodesic equation . the action ( [ 1 ] ) is the first order form of the better known second order action . the latter can be obtain from ( [ 1 ] ) by solving the momentum equation of motion and substituting it back to ( [ 1 ] ) ; as a result one obtains @xmath14 before we proceed , let us recall the basic properties of the tetrads . it follows from the defining equation that tetrads are spacetime vectors , transforming under diffeomorphisms as @xmath15 moreover the relations ( [ 1a ] ) are invariant under infinitesimal local lorentz transformations @xmath16 these two symmetries commute @xmath17\ , e^a_\mu=0 $ ] . in addition , assuming vanishing torsion , the tetrad satisfies @xmath18 } + \omega_{[\mu}{}^a{}_{b}\ , e^b_{\nu]}=0\,,\ ] ] where @xmath19 is a gauge field for local lorentz symmetry , transforming as @xmath20 where @xmath21 is the matrix of a finite lorentz transformation @xmath22 , with @xmath23 being the matrix generators of the lorentz group . finally the tetrads satisfy the tetrad postulate , according to which they are covariantly constant @xmath24 it follows from the the born reciprocity idea that the kinetic part of the relativistic particle lagrangian in the case of curved momentum space should look like ( [ 1 ] ) with the roles of @xmath25 and @xmath13 exchanged . one also has to replace the mass - shell condition with its curved momentum space counterpart @xmath26 , where @xmath27 is the square of the geodesic distance between the point with coordinates @xmath28 and the origin of the momentum space , obtaining as a result @xcite , @xcite @xmath29 it should be recalled at this point that although the action ( [ 2 ] ) looks much more complex than the one of the standard relativistic particle moving in flat space , the equations of motion following from it are remarkably similar : they say that momentum @xmath30 and velocity @xmath31 are both constant . what makes the action ( [ 2 ] ) different from its standard counterpart is the relation between momentum and velocity , which becomes in the case of ( [ 2 ] ) highly nonlinear . moreover , when interactions between particles are introduced , the effects of relative locality starts being visible . generalizing these considerations , we want now to construct an action for a particle with momentum space and spacetime both possessing nontrivial geometries . of course , we want the action to reproduce the two limiting cases of flat momentum space / flat spacetime discussed above . we require moreover the new action to be still manifestly invariant under general coordinate transformations . in order to meet these requirements , we introduce non local variables which we denote @xmath32 . we will describe their construction in the following subsection . let us denote by @xmath33 the @xmath34 curve @xmath35 for @xmath36 $ ] and consider the sub - curves @xmath37 with @xmath38 $ ] which coincide with @xmath33 up to @xmath39 . as discussed above , given the background spacetime metric @xmath40 , there exist a whole family of tetrads satisfying eq.([1a ] ) , which differ from one another by action of local lorentz transformations . we can now gauge fix the local lorentz transformations in such a way that the lorentz connection @xmath41 vanishes along a given curve @xmath35 and then it follows from ( [ 1d ] ) that on this curve we can construct a tetrad @xmath42 with @xmath43 such that @xmath44 the existence of the lorentz connection with these properties can be proved as follows . we first show that the component of @xmath45 along the worldline @xmath33 , @xmath46 can be gauge fixed to zero . to this end we have to solve the equation ( cf . ( [ 1e ] ) ) @xmath47 . but this equation is solved by a time ordered wilson line ( holonomy ) , @xmath48 . having fixed @xmath49 we are left with the gauge transformations that are constant along @xmath33 . let us now consider an arbitrary constant time surface , corresponding to some particular value of the parameter @xmath5 on @xmath33 . then in the vicinity of the point in which the worldline crosses the surface we have @xmath50 . this is sufficient freedom to gauge fix to zero the spacial components of lorentz connection @xmath51 . this can be done for any @xmath5 and therefore all the components of lorentz connection can be gauge fixed to zero along a curve . after gauge fixing @xmath52 on @xmath33 , we are left with the gauge freedom @xmath53 . therefore , once we gauge fix the connection along one curve , in general we can not do the same for another curve in its small neighborhood . the tetrads @xmath42 ( [ 3 ] ) are determined modulo a global lorentz transformation @xmath54 , which can be used to fix them equal to an arbitrary tetrad @xmath7 at one point of @xmath33 , @xmath55 since the construction of the tetrads @xmath42 reminds the one of fermi coordinates , in what follows we will call them fermi tetrads . with these prerequisites we are ready to define the nonlocal variable @xmath32 as @xmath56 the variable @xmath57 depends in general on the curve @xmath33 along which it is calculated . and @xmath58 . ] in order to calculate the variation of the action below , we will have to evaluate the difference between variables @xmath32 calculated along different curves @xmath33 and @xmath59 lying infinitesimally close to each other , with appropriate fermi tetrads associated with each of them . thus @xmath60 where @xmath61 denotes a tetrad field , defined in spacetime , which becomes a fermi tetrad on the curve @xmath33 . the variation of the tetrads @xmath42 can be decomposed into two parts : @xmath62 with @xmath63 and @xmath64 in order to evaluate the expression ( [ delta1 ] ) we need to compare the fermi tetrads associated with different curves . since they are both tetrads of the same spacetime metric , there exist a local lorentz transformation @xmath21 relating them @xmath65 and the associated lorentz connections are related by the local lorentz gauge transformation @xmath66 we know that the lorentz connection @xmath67 vanishes on @xmath59 and @xmath68 vanishes on @xmath33 ; thus we have @xmath69 for an infinitesimal local lorentz gauge transformation @xmath70 , keeping the leading order terms we obtain @xmath71 from which , multiplying by @xmath72 and adding a term proportional to @xmath73 , we get @xmath74 where in the last line we used an expression for the curvature tensor that holds on @xmath33 @xmath75 equation ( [ 19 ] ) is solved by @xmath76 where we used again the fact that @xmath77 vanishes on @xmath33 and @xmath78 hence , the total variation reads @xmath79 where we used @xmath80 . notice that when @xmath81 , @xmath82 vanishes and one gets @xmath83 the total variation can be rewritten via an integration by part as @xmath84 this expression provides a linear map between the variations @xmath85 and @xmath86 . since the basic physical variable of the particle model is the position of the worldline @xmath4 , and because the expression ( [ varfin ] ) is very complex and nonlocal , to make sure that the equations of motion following from varying @xmath87 and @xmath4 are identical , we must show that the linear mapping ( [ varfin ] ) is invertible . this is done in the appendix a. equations ( [ 7 ] ) and ( [ varfin ] ) contain all the information we need to construct the action of a free particle moving in curved spacetime and momentum space , and compute the corresponding equations of motion from the variational principle . using the non - local variable @xmath32 discussed in the preceding subsection we define the action of a particle moving in curved spacetime and momentum spaces as follows @xmath88\ , e_a^\alpha\ , \dot{p}_\alpha\,+n({\cal c}(p)\,-\,m^2)\right\}\,,\ ] ] where @xmath89 $ ] is calculated along the same curve @xmath33 as the integral in ( [ 9 ] ) . before turning to the discussion of the properties of this action it is worth checking if it acquires the desired form in the limiting cases of flat spacetime / momentum space , respectively . in the case of flat spacetime , to evaluate ( [ 5 ] ) we have to find the form of the associated fermi tetrad . it follows from the tetrad postulate ( [ 1f ] ) that since both @xmath90 and @xmath91 vanish such tetrad must satisfy @xmath92 , and thus @xmath93 ( up to a global lorentz transformation . ) then @xmath94 it can be checked that the second ( constant ) term produces neither a contribution to the equation of motion nor a boundary term , and thus for flat spacetime the action reproduces the one of relative locality . in the opposite case , when the momentum space is flat , @xmath95 . we integrate ( [ 9 ] ) by parts and use ( [ 7 ] ) to obtain the standard curved spacetime particle action ( up to a boundary term ) , with the only difference being that now we have to do with the fermi tetrad instead of the generic one . however , the equations of motion are the same in both cases , so we may conclude that the actions are equivalent , the only difference being that the lagrangian ( [ 1 ] ) is invariant under local lorentz symmetry , while in the action ( [ 9 ] ) only the global lorentz symmetry remains ) differs , up to a boundary term , from the action presented in @xcite ( written for fermi tetrads ) by the term @xmath96 . ] . now we can turn to the equations of motion following from the action ( [ 9 ] ) . its variation reads @xmath97 since we demonstrated that the map ( [ varfin ] ) is invertible , we know that the equations of motion we get from the stationarity of the action ( @xmath98 ) with respect to arbitrary variations @xmath85 are equivalent to the ones obtained by considering arbitrary variations @xmath86 . hence , we get @xmath99 which are equivalent to ( for constant @xmath9 ) @xmath100 in particular , from the second relation in ( [ 12 ] ) one finds @xmath101 where we used the expression of the christoffel symbols in terms of the tetrads @xmath42 , that follows from the tetrad postulate ( [ 1f ] ) @xmath102 therefore , the trajectory in spacetime is a geodesic , independently of the geometry in momentum space . having discussed the form of the action , let us now consider its symmetries . first of all the action is manifestly invariant under general coordinate transformations , in both momentum space ) , the invariance under momentum space diffeomorphisms , which could be a mere formal invariance , has no classical counterpart , and we postpone its discussion to future studies . ] and spacetime , since @xmath32 is a spacetime scalar . second it is invariant under residual , global lorentz transformations that leave invariant the condition that the connection vanishes along the curve @xmath103 . from the relative locality perspective we are especially interested in translational symmetries : in the case of the model of a particle moving in flat spacetime , the main features of relativity of spacetime locality are encoded in that the fact that the translations become momentum - dependent @xcite , @xcite , @xcite . as we will see an analogous effect takes place in the case of curved spacetime . like in the case of relative locality in flat spacetime @xcite , @xcite , @xcite , the action ( [ 9 ] ) is invariant ( up to a boundary term ) under the translation @xmath104 in flat - spacetime relative locality this symmetry translates rigidly a ( straight ) particle worldline by an amount that depends on the momentum carried by the particle . as we will see , an analogous effect takes place in curved spacetime . to see this we must find out what is the infinitesimal shift of the particle trajectory @xmath85 corresponding to the translation ( [ 30 ] ) . since we are interested in the effect that the transformation ( [ 30 ] ) has on trajectories , we assume that the equations of motion are satisfied . we start with ( [ varfin ] ) @xmath105 at @xmath106 we have @xmath107 this defines the first initial condition for @xmath108 . next , let us differentiate ( [ symm2 ] ) over @xmath5 @xmath109 so that at @xmath110 @xmath111_{t_1}=0\,.\ ] ] taking the second derivative of ( [ symm2 ] ) over @xmath5 we find @xmath112 since @xmath113 is zero by equations of motion , the second term in the above expression disappears and we are left with @xmath114 as shown in appendix b this equation can be rewritten as @xmath115 where @xmath116 is the covariant derivative projected along the worldline , subject to the initial conditions @xmath117 equation ( [ symm10 ] ) is an equation of geodesic deviation and therefore we see that the translational symmetry ( [ 30 ] ) maps the original geodesic , being the particle worldline , to another one , with the magnitude of translation depending on the momentum carried by the particle . this is exactly the effect one could foresee from the flat spacetime relative locality , where straight lines ( geodesics ) are translated by a constant , momentum dependent amount . it follows from ( [ symm11 ] ) that @xmath85 has the momentum dependence encoded by the initial condition . let us define another variable @xmath118 , which describe the momentum independent translation @xmath119 since , as shown in appendix b , both first and second covariant derivatives of the tetrad @xmath120 along the worldline vanish , we can rewrite ( [ symm10 ] ) as @xmath121 this equation describes a congruence of particle worldlines in the spacetime whose curvature is momentum - dependent . it might serve as a starting point of more phenomenologically oriented investigations . in the spirit of the relative locality framework introduced in @xcite ( see also @xcite for an extensive discussion of the properties of these boundary terms ) , in order to describe particle processes ( at a semi - classical , non quantum , level ) , we introduce in the action ( [ 3 ] ) boundary terms enforcing constraints on the endpoints of the particles worldlines . to illustrate how such a constraint may be introduced in our framework , suppose that we want to describe an idealized process depicted in fig . [ vertexfig ] , with two incoming particles labeled respectively by momenta and coordinates @xmath122 , @xmath123 and the outgoing one labeled by @xmath124 . , @xmath125 and one outgoing particle of momentum @xmath126 . @xmath127 is a function of the particles momenta enforcing energy - momentum conservation at the vertex . the graphic must not be intended as a spacetime representation but just as a qualitative picture illustrating the combination of momenta in the process . ] the action for this process is @xmath128 \nonumber\\ + & \int_{-\infty}^{t}d\tau\left [ y^{a}e_a^\alpha\dot{q}_{\alpha } + n_q\left({\cal c}(q ) - m_q^{2}\right)\right]\nonumber\\ + & \int^{\infty}_{t}d\tau\left[z^{a}e_a^\alpha\dot{r}_{\alpha } + n_r\left({\cal c}(r ) - m_r^{2}\right ) \right]- k^{\alpha}{\cal k}_{\alpha}\left(p , q , r\right ) \bigg\|_{\bar{\tau}}\ , . \label{actionint}\end{aligned}\ ] ] in this formula @xmath129 is a lagrange multiplier enforcing the constraint @xmath130 on the worldlines endpoints , which plays a role of a ( in general ) deformed law of energy - momentum conservation at the vertex . typically , @xmath131 , with the symbol @xmath132 ( @xmath133 ) encoding the connection on the momentum space geometry , characterizing the ( in general non - linear ) law of summation for momenta @xcite . in addition to the equations of motion for the bulk part eqs . ( [ 11 ] ) , ( [ 13 ] ) , the boundary term contributes with the constraints @xmath134 when @xmath129 changes , the @xmath32 transforms as @xmath135 with analogous relations holding for the other particles . assuming that we take the initial condition for the geodesic deviation equation ( [ symm11 ] ) at the interaction point , we find that @xmath136 we see therefore that the structure of the interaction vertex in the case of curved spacetime is essentially the same as in the flat spacetime case of relative locality @xcite . in this paper we generalized relative locality , originally defined @xcite in flat spacetime , to the case of an arbitrary , curved background spacetime , preserving invariance under general coordinate transformations . it turns out that on the formal level this latter theory is a natural generalization of the former one : free particle trajectories are now geodesics instead of flat - spacetime straight lines , and rigid , momentum dependent translations of the flat case are replaced with geodesic deviations , sensitive to curvature . in spite of the apparent similarities there are , however , some major differences between flat and curved spacetime case . in the latter we were forced to use nonlocal variables @xmath32 to define the action that had desired symmetry properties . this might be just a technical artifact , but it may also signal a presence of some deeper layer present in theories with a nontrivial geometry in both momentum space and spacetime . furthermore , in flat spacetime the symmetries of the action are associated with some transformations ( rigid translations ) defined in the whole spacetime manifold . on the contrary , in the curved case we have to solve the equation of geodesic deviation on a given geodesic to find the symmetry . this implies that the transformation @xmath85 which leaves the action invariant ( up to a boundary term ) depends on each particular solution of equations of motion and generically , because of curvature , it can not be extended to the whole spacetime . aside from its conceptual relevance , our result opens some interesting phenomenological perspectives . indeed most of the opportunities to test planck - scale deformation effects on particle kinematics , that have been proposed in the recent literature , rely on some source of amplification of the relevant effects due to cosmological distance of astrophysical sources . most of the results for theories with curved momentum space ( and earlier of the dsr theories ) have been discussed in the context of flat spacetime , while in the proposed scenarios relevant for planck - scale phenomenology the effect of spacetime curvature can not be neglected . our result can be taken then as starting point for further studies of relative locality effects in presence of spacetime curvature . moreover using the results of the present work one may try to investigate the relative locality effects in the case of strong gravitational field , for example in the context of black hole physics ( see @xcite ) . we would like to thank laurent freidel and lee smolin for discussion . for fc , jkg , and gr this work was supported by funds provided by the national science center under the agreement dec- 2011/02/a / st2/00294 . for jkg this work was also supported in parts by the grant 2011/01/b / st2/03354 . to show that the map ( [ varfin ] ) is invertible we must show that its kernel contains only @xmath137 , to wit @xmath138 in order to prove this let us note that since we assumed @xmath139 to be a @xmath34 function of @xmath5 , @xmath87 and @xmath140 are @xmath34 too and the condition @xmath141 for each @xmath5 is equivalent to @xmath142\bigg\|_0=0,\qquad \forall n\in\mathbb{n}\,.\ ] ] from the expression ( [ varfin ] ) we have @xmath143 and for @xmath144 we get @xmath145\bigg\|_0=\bigg[\frac{d}{d\tau}\left(\bar{e}_{\nu}^a\,\delta x^\nu\right)\bigg]\bigg\|_0\,,\ ] ] which implies ( we remind that @xmath146 ) @xmath147\bigg\|_0=0\,.\ ] ] similarly , one can show that @xmath148\bigg\|_0=0\rightarrow\bigg[\frac{d^2}{d\tau^2}\delta x^\mu\bigg]\bigg\|_0=0\,,\ ] ] and by iterating one gets @xmath149\bigg\|_0=0,\qquad \forall n\in\mathbb{n}\,,\ ] ] which is equivalent to @xmath150 for each @xmath36 $ ] . therefore we proved ( [ 0kernel ] ) and the map ( [ varfin ] ) is invertible . in this appendix we show that eq . ( [ symm7 ] ) is equivalent to the equation of geodesic deviation ( [ symm10 ] ) . to see this notice that @xmath151 and @xmath152 it follows from the tetrad postulate ( [ 1f ] ) and the properties of fermi tetrads that @xmath153 as for the second covariant derivative we get @xmath154 this expression is again zero for a fermi tetrad , because connection @xmath19 is zero everywhere on the worldline and thus its derivative along it vanishes ; therefore @xmath155 using ( [ b1 ] ) and ( [ b2 ] ) one straightforwardly derives ( [ symm10 ] ) . s. majid , `` meaning of noncommutative geometry and the planck scale quantum group , '' lect . notes phys . * 541 * , 227 ( 2000 ) [ hep - th/0006166 ] . h. -j . matschull and m. welling , `` quantum mechanics of a point particle in ( 2 + 1)-dimensional gravity , '' class . grav . * 15 * ( 1998 ) 2981 [ gr - qc/9708054 ] . l. freidel and e. r. livine , `` effective 3-d quantum gravity and non - commutative quantum field theory , '' phys . * 96 * ( 2006 ) 221301 [ hep - th/0512113 ] . g. amelino - camelia , `` testable scenario for relativity with minimum - length , '' phys . b * 510 * , 255 ( 2001 ) [ arxiv : hep - th/0012238 ] . g. amelino - camelia , `` relativity in space - times with short - distance structure governed by an observer - independent ( planckian ) length scale , '' int . j. mod . phys . d * 11 * , 35 ( 2002 ) [ arxiv : gr - qc/0012051 ] . j. kowalski - glikman , `` observer independent quantum of mass , '' phys . a * 286 * ( 2001 ) 391 [ hep - th/0102098 ] . j. kowalski - glikman and s. nowak , `` doubly special relativity and de sitter space , '' class . grav . * 20 * ( 2003 ) 4799 [ hep - th/0304101 ] . g. amelino - camelia , m. matassa , f. mercati , g. rosati , `` taming nonlocality in theories with planck - scale - deformed lorentz symmetry '' , phys . * 106 * ( 2011 ) 071301 , [ arxiv:1006.2126 [ gr - qc ] ] . g. amelino - camelia , l. freidel , j. kowalski - glikman and l. smolin , `` the principle of relative locality , '' phys . d * 84 * ( 2011 ) 084010 [ arxiv:1101.0931 [ hep - th ] ] . g. amelino - camelia , l. freidel , j. kowalski - glikman and l. smolin , `` relative locality : a deepening of the relativity principle , '' gen . * 43 * ( 2011 ) 2547 int . j. mod . d * 20 * ( 2011 ) 2867 [ arxiv:1106.0313 [ hep - th ] ] . g. amelino - camelia , a. marciano , m. matassa and g. rosati , `` deformed lorentz symmetry and relative locality in a curved / expanding spacetime , '' phys . d * 86 * ( 2012 ) 124035 [ arxiv:1206.5315 [ hep - th ] ] . j. kowalski - glikman and g. rosati , `` relative locality in curved space - time , '' mod . a * 28 * ( 2013 ) 1350101 [ arxiv:1303.7216 [ hep - th ] ] . l. freidel and t. rempel , `` action and vertices in the worldine formalism , '' arxiv:1312.5396 [ hep - th ] . g. amelino - camelia , m. arzano , j. kowalski - glikman , g. rosati and g. trevisan , `` relative - locality distant observers and the phenomenology of momentum - space geometry , '' class . quant . * 29 * ( 2012 ) 075007 [ arxiv:1107.1724 [ hep - th ] ] . j. kowalski - glikman , `` living in curved momentum space , '' int . j. mod . a * 28 * ( 2013 ) 1330014 [ arxiv:1303.0195 [ hep - th ] ] . l. smolin , `` the black hole information paradox and relative locality , '' arxiv:1108.0910 [ gr - qc ] .	we construct a theory of particles moving in curved both momentum space and spacetime , being a generalization of relative locality . we find that in order to construct such theory , with desired symmetries , including the general coordinate invariance , we have to use non local position variables . it turns out that free particles move on geodesics and momentum dependent translations of relative locality are replaced with momentum dependent geodesic deviations .
transition metal dichalcogenides ( tmdcs ) are layered materials with the chemical composition mx@xmath1 , where m is a transition metal element ( such as molybdenum or tungsten ) , and x is a chalcogen ( sulfur , selenium , or tellurium ) . the interest to semiconducting tmdcs has been sparked by the recent discovery of the monolayer mos@xmath1 being a direct - gap semiconductor , in contrast to its bulk indirect - gap counterpart @xcite . atomically thin tmdc monolayers can be extracted from bulk crystals by exfoliation , similarly to graphene @xcite , or grown by molecular beam epitaxy or chemical vapor deposition @xcite . the optical gap in the visible light range @xcite and tightly bound excitons @xcite , make them quite promising for optical applications @xcite . a unique feature of tmdcs is the so - called spin - valley locking @xcite . the conduction and valence band extrema are located at the two inequivalent @xmath2 points ( valleys ) of the hexagonal first brillouin zone . absence of inversion symmetry and strong spin - orbit interaction , originating from @xmath3-orbitals of the metal atoms , leads to spin splitting of the bands . the sign of the splitting is opposite in the two valleys , as required by the time - reversal symmetry , so the lowest - energy electron and hole states have opposite spin projections in the opposite valleys , as shown schematically in fig . [ fig : tmdcbands ] . it is important that the two valleys can be addressed separately by optical means . namely , in each valley optical transitions with only one of the two in - plane circular polarizations are allowed @xcite . this opens an exciting perspective of manipulating the spin / valley degree of freedom optically @xcite . ( color online . ) a schematic view of the conduction and valence band extrema of a tungsten - base tmdc monolayer . the solid arrows show the spin projections . in molybdenum - based compounds , the spin projections in the conduction band are the opposite . the dashed arrows represent circularly polarized optical transitions for @xmath4 excitons in each valley . , width=302 ] the large spin splitting in the valence band ( a fraction of ev ) gives rise to two prominent features in the absorption spectrum , called @xmath4 and @xmath5 excitons , formed by a hole in the upper or lower spin - split valence band , respectively , and an electron in the conduction band . the spin splitting in the conduction band is much weaker ( on the mev scale ) ; moreover , its sign depends on the material : in molybdenum - based compounds , the top valence band and the lowest conduction band have the same spin projection in the same valley , while in tungsten - based ones , they are opposite ( fig . [ fig : tmdcbands ] ) @xcite . this implies that the lowest - energy @xmath4 excitons in wx@xmath1 involve electronic states either from the same valley but with opposite spins , or with the same spin but from different valleys . as the photons , polarized in the plane , couple only to the orbital degrees of freedom and carry a very small momentum , they are decoupled from such excitons , which therefore have been named `` dark '' . this picture provided a natural explanation of the experimentally observed rise of the luminescence intensity in wse@xmath1 with increasing temperature in terms of the increasing thermal population of the higher - energy bright @xmath4 excitons @xcite . at sufficiently low temperatures , several peaks at energies below the neutral and charged bright @xmath4 excitons have been seen in the wse@xmath1 luminescence spectrum @xcite . these observations pose an important question about existence of possible mechanisms for radiative decay of the dark excitons . while the radiative decay of bright excitons in tmdcs fits in the standard picture for 2d excitons @xcite and has been described in several theoretical works @xcite , the dark exciton radiative decay requires unlocking the spin and the valley degrees of freedom . namely , for intervalley dark excitons , a large momentum should be provided to the electron or to the hole without flipping its spin , which requires a third body ( a phonon , an impurity , another electron or a hole ) . for intravalley dark exciton decay , a spin flip is required . investigation of this latter possibility is the subject of the present work . in this paper , we give an estimate of the intravalley dark exciton radiative decay rate due to several spin - flip mechanisms . one such mechanism could be the bychkov - rashba effect @xcite , as mentioned in ref . it requires breaking of the crystal symmetry with respect to reflection in the monolayer plane . this can be achieved by applying an external electric field , perpendicular to the plane ; the corresponding rashba coefficient has been calculated in ref . the reflection symmetry can also be broken if the dielectric environments on the two sides of the monolayer are different . having not found any information on the magnitude of this effect in the literature , we make our own crude estimate ; for a tmdc monolayer on a glass - like substrate with the empty space above , we obtain an equivalent electric field to be roughly of the order of 0.1 v / . however , even for this relatively high field , the resulting radiative rate is about @xmath6 of the bright exciton decay rate @xmath7 , which makes the rashba mechanism totally negligible with respect to other relaxation processes . the main reason for such smallness is that rashba spin - flip amplitude is proportional to the electron momentum , and radiative decay is possible only in a narrow radiative region of small momenta . another way to flip the electron spin is to apply an in - plane magnetic field and use the zeeman effect . it is momentum - independent , and for a quite high but still realistic field of 30 t we obtain a radiative rate exceeding @xmath8 . finally , we analyzed the intrinsic radiative decay mechanism due to the interband spin - flip dipole moment perpendicular to the monolayer plane , mentioned in refs . in fact , it is the presence of this dipole moment that gives rise to bychkov - rashba coupling in a perpendicular electric field @xcite , so its magnitude can be deduced from the estimates of ref . @xcite for the rashba coefficient . this enabled us to estimate the associated radiative decay rate as @xmath9 . however , we also found that the same out - of - plane interband dipole gives rise to a coulomb local - field term which lifts the double degeneracy between the dark excitons made of electronic states from the two valleys . this local - field effect is analogous to the exchange energy shift of the @xmath10 excitons in semiconductor quantum wells @xcite , and produces an energy splitting which we very roughly estimate as about 10 mev . its precise evaluation requires a microscopic treatment on the atomic scales , similar to that in ref . crucially , the whole oscillator strength of the interband spin - flip dipole goes into the higher - energy component . thus , we find that the dark intravalley @xmath4 exciton has , in fact , two components , one really dark and the other one about 1001000 times darker than the bright exciton , so it can be called `` dim '' . because of the splitting , at low temperatures the emission from the dim exciton should be suppressed by an activation factor , analogously to that from the bright one . note that the dark component can be made decay by applying an in - plane magnetic field , as mentioned above , and the corresponding rate can be made comparable to that of the dim component for a sufficiently high field . the paper is organized as follows . in sec . [ sec : model ] we describe the effective model for electrons in the tmdc monolayer , the excitonic states , and specify how different spin - flip mechanisms enter the model . in sec . [ sec : polarization ] , we discuss spin - mixing of excitonic states and calculate the `` mechanical '' susceptibilities of the tmdc monolayer , which determine the exciton coupling to the macroscopic electromagnetic field . in sec . [ sec : radiative ] we study the effect of the exciton coupling to the electromagnetic field and compute the radiative energy shifts and decay rates for the excitons . finally , in sec . [ sec : conclusions ] we summarize our results and discuss some of their implications and perspectives . some details of calculations are presented in several appendices . we set @xmath11 throughout the paper , and write the electronic hamiltonian of a tmdc monolayer as @xmath12 its three terms will be discussed in the following subsections . we assume the monolayer to be in the @xmath13 plane , the three - dimensional position vector @xmath14 represented @xmath15 , so that @xmath16 denotes the position in the monolayer plane . the first term in eq . ( [ h= ] ) , @xmath17 , is the usual effective two - band hamiltonian @xcite @xmath18 written in terms of four - component column operators @xmath19^t$ ] , where @xmath20 labels the two valleys @xmath2 , and @xmath21 is the @xmath22 matrix , written in the block form , @xmath23.\ ] ] here @xmath24 is the third pauli matrix in the spin subspace , @xmath25 is the band gap , @xmath26 is the velocity matrix element between the band extrema bloch functions ( which can be made real by an appropriate choice of the relative phase between the conduction and valence band bloch functions ) . the coefficients @xmath27 are related to the electron and hole effective masses @xmath28 as @xmath29 typically , @xmath30 , @xmath31 , @xmath32 for mox@xmath1 and @xmath33 for wx@xmath1 ( @xmath34 being the free electron mass ) . @xmath35 in eq . ( [ htauk= ] ) is half of the spin splitting in the conduction / valence bands . we distinguish the @xmath2 valleys by assuming the valence band states to originate predominantly from the metallic orbitals with the @xmath36 projections of the angular momentum equal to @xmath37 at the @xmath2 point , respectively . this fixes @xmath38 , as well as the valley - dependent optical selection rules : the left / right circular polarisation can be absorbed in the @xmath2 valley , respectively , as the conduction band states originate mostly from zero - angular - momentum metallic orbitals and the angular momentum is conserved modulo 3 due to the three - fold crystal rotation symmetry . typically , @xmath39 in mox@xmath1 and @xmath40 in wx@xmath1 . @xmath41 is much weaker , usually a few tens of mev , with the exception of mos@xmath1 where it is extremely weak , @xmath42 . the sign of @xmath41 depends on the material : @xmath43 ( @xmath44 ) in molybdenum - based ( tungsten - based ) compounds . because of this , the lowest - energy interband transition in tungsten - based compounds involves either a spin flip or valley switching . the second term in eq . ( [ h= ] ) , @xmath45 , represents the coulomb interaction between electrons , @xmath46 \hat\psi_\tau({\mathbf{r } } ) . \label{rhomatrix=}\end{aligned}\ ] ] here each element of the block @xmath47 matrix should be understood as the unit matrix in the spin space . we included only the intra - valley part of the electron density @xmath48 , as it is the long - range part of the coulomb interaction ( on the scale of lattice constant ) which is responsible for the exciton formation . the contribution of the diagonal matrix elements in eq . ( [ rhomatrix= ] ) corresponds to the direct coulomb attraction between the electron and the hole , which gives rise to the excitonic bound state . the wave function of the electron - hole relative motion , @xmath49 , is obtained from the corresponding schrdinger equation , @xmath50\phi({\mathbf{r}})= e\,\phi({\mathbf{r}}),\ ] ] where @xmath51 is the reduced mass . the lowest eigenvalue @xmath52 of eq . ( [ excschrodinger= ] ) defines the exciton binding energy @xmath53 . as we are working in the parabolic approximation for the electronic dispersion and assume the electron and the hole masses to be spin - independent , the wave function @xmath54 and the binding energy @xmath53 are the same for all spin and valley configurations of the electron and the hole . we do not specify the explicit form of the pair interaction potential @xmath55 , as we will not be solving eq . ( [ excschrodinger= ] ) . it is known that because of strong dielectric confinement in the tmdc monolayer , the interaction potential is not @xmath56 , so the bound state wave functions do not have a hydrogenic form @xcite . instead , we will assume @xmath54 and @xmath53 to be known and treat them as input parameters . in fact , we will not need the whole wave function @xmath54 , but only its value @xmath57 at the coinciding electron and hole positions . by the order of magnitude , @xmath57 is the inverse radius of the excitonic bound state . estimating it as @xmath58 with the binding energy @xmath59 @xcite , gives @xmath60 , while from the measured diamagnetic shift of the exciton energy one infers values slightly below @xmath61 @xcite . the contribution from the off - diagonal matrix elements in eq . ( [ rhomatrix= ] ) corresponds to the exchange interaction between the electron and the hole . it will be taken into account in sec . [ sec : radiative ] in the framework of macroscopic electrodynamics . the third term in eq . ( [ h= ] ) , @xmath62 , is the spin - flip , which arises when the reflection symmetry in the @xmath36 direction is broken @xcite . we write this term as @xcite @xmath63,\end{aligned}\ ] ] here we assumed that the reflection symmetry is broken by an external perpendicular electric field @xmath64 , introduced explicitly . then , by definition , the operator multiplying @xmath64 is nothing but the @xmath36 component of the electric dipole moment operator . if there is no external electric field , but the reflection symmetry is broken by some other mechanism ( e. g. , van der waals interaction with a dielectric substrate ) , the coupling hamiltonian still has the form ( [ hrashba= ] ) which is fixed by the symmetry , but instead of @xmath65 its strength is determined by another parameter having the dimensionality of energy . for example , the parameter @xmath66 from ref . @xcite is related to @xmath65 as @xmath67 . note that the parameter @xmath68 can be made purely real or purely imaginary by an appropriate spin rotation around the @xmath36 axis . the peculiar valley structure of the matrix @xmath69 follows from the total ( orbital plus spin ) angular momentum conservation : in @xmath70 ( @xmath71 ) valley only spin - up ( spin - down ) states in the valence band can couple to only spin - down ( spin - up ) in the conduction band , the corresponding total angular momentum change being @xmath72 . the hamiltonian @xmath62 produces interband spin - flip transitions . combining it with the off - diagonal terms in eq . ( [ htauk= ] ) in the first - order perturbation theory , one can project the spin - flip terms on the conduction and valence bands ( appendix [ app : diag_hamiltonian ] ) . upon projection , they acquire the familiar rashba - type form @xcite , linear in momentum @xmath73 : @xmath74_z{\mathop{\mathrm{im}}}{d}_z\right\ } \left[\begin{array}{cc } 1 & 0 \\ 0 & -1 \end{array}\right].\ ] ] this form matches the one studied in ref . @xcite where the rashba coupling strength was parametrized by @xmath75 , related to @xmath65 as @xmath76 . the parameter @xmath75 was estimated in ref . @xcite for mos@xmath1 , mose@xmath1 , ws@xmath1 , wse@xmath1 monolayers suspended in vacuum and subject to an external perpendicular electric field . this enables us to extract @xmath77 of about @xmath78 , @xmath79 , @xmath80 , and @xmath81 , for mos@xmath1 , mose@xmath1 , ws@xmath1 , and wse@xmath1 , respectively . we are not aware of any estimates in the literature for the rashba coupling strength induced by a dielectric substrate . therefore , we make our own crude estimate of the effective electric field in appendix [ app : substrate ] , which gives @xmath82 for a wx@xmath1 monolayer lying on a glass - like substrate with a dielectric constant @xmath83 and the empty half - space above . finally , we note that in addition to rashba effect in a perpendicular electric field , the spin can be flipped by the zeeman effect if an in - plane magnetic field @xmath84 is applied . this effect is taken into account straightforwardly by adding @xmath85 to the diagonal matrix elements in eq . ( [ htauk= ] ) , where @xmath86 is the bohr magneton and @xmath87 is the in - plane gyromagnetic ratio for electrons in the conduction and valence band . the hamiltonian of interaction between the electrons and the transverse photons , which are described by the long - wavelength vector potential @xmath88 , is usually obtained from the requirement of gauge invariance by performing the peierls substitution in eq . ( [ htauk= ] ) : @xmath89 . however , besides coupling to photons via gauge potentials , electrons can couple directly to electric and magnetic fields . such terms are gauge invariant and can not be deduced from the bare electronic hamiltonian in the envelope function approximation , as they correspond to the effect of electromagnetic field on the microscopic bloch functions . in the previous subsection , we have already presented such terms , corresponding to rashba and zeeman effects . it is crucial for our analysis that , once the dipole operator in eq . ( [ hrashba= ] ) is determined , it can also describe the electronic coupling to the photon field . namely , @xmath64 can be understood not just as a static external field , but as the electric field of photons as well . in this case , one should write @xmath90 and put it inside the space integral . the corresponding hamiltonian can then describe interband photon absorption / emission , accompanied by the spin flip ( the existence of such coupling was briefly mentioned in ref . the main quantity which we use to describe the interaction of excitons with light is the susceptibility @xmath91 , defined as the mechanical response of the excitonic polarization to the electric field at the frequency @xmath92 and the in - plane wave vector @xmath93 . ( the term `` mechanical '' implies that the only the direct coulomb interaction is taken into account for the moment ; the exchange interaction will be included in the framework of macroscopic electrodynamics . ) we define the polarization @xmath94 as the dipole moment per unit area . it is a three - dimensional vector , whose @xmath36 component is defined by eq . ( [ hrashba= ] ) , while the in - plane components are most conveniently defined via the in - plane current @xmath95 . the current operator is obtained in the usual way from @xmath96 , where we retain only the leading terms at @xmath97 : @xmath98 \hat\psi_\tau({\mathbf{r}}),\ ] ] where @xmath99 are the unit vectors in the corresponding directions . the kubo formula gives [ chidef=]@xmath100 where @xmath101 and @xmath102 are the ground and excited electronic states , @xmath103 is their energy difference ( that is , the energy of the excitation @xmath104 ) , the indices @xmath105 , and we omitted the non - resonant terms . in the translationally invariant case , the susceptibilities depend only on the difference @xmath106 , so @xmath107 here the in - plane components @xmath108 are labeled by the greek indices @xmath109 , while the latin indices run over all three dimensions , @xmath110 . in the following , we evaluate @xmath91 using the eigenstates @xmath102 which are the lowest - energy states of the intravalley excitons , whose spins are mixed by the rashba coupling . the effect of the zeeman coupling is treated analogously , the result is summarized in the end of sec . [ ssec : susceptibility ] . we start by constructing the zero - approximation states , i. e. , those in the absence of the rashba coupling . first , each intravalley exciton can be characterized by the index @xmath111 , indicating the valley in which the electronic transition takes place . it gives rise to two species of intravalley excitons , which can be distinguished by the valley index @xmath20 . further , in each valley the conduction and valence bands are spin split , so the excitonic states can be labeled by a pair of spin indices @xmath112 referring to the conduction and the valence band , respectively ( the hole spin is given by @xmath113 ) . finally , the excitonic state is characterized by a center - of - mass momentum @xmath93 . so , the zero - approximation excitonic states are written as @xmath114 here , @xmath115 is the sample area , @xmath116 is the center - of - mass coordinate , @xmath117 , and @xmath49 is the normalized wave function of the relative electron - hole motion , corresponding to the lowest bound state , which was discussed in sec . [ ssec : coulomb ] . states ( [ state0= ] ) are normalized as @xmath118 and have the energies @xmath119 where we introduced the excitonic mass @xmath120 . the rashba coupling flips electron and hole spins and mixes the excitonic states with different @xmath121 . as @xmath122 , we neglect the spin - flip in the valence band , and consider it only in the conduction band . it can be described by an effective @xmath47 hamiltonian in the basis of @xmath123 and @xmath124 , obtained by the projection of the electronic rashba hamiltonian ( [ rcvb= ] ) on the excitonic states : @xmath125,\ ] ] where the excitonic rashba coupling constant is given by @xmath126 description of spin flip in terms of the purely excitonic effective hamiltonian ( [ rashbaex= ] ) assumes that the spin flip does not disturb the electron - hole relative motion , which is guaranteed by the condition @xmath127 @xcite . the @xmath47 hamiltonian ( [ rashbaex= ] ) can be diagonalized exactly ; however , we are interested only in small momenta @xmath128 , so perturbative expressions will suffice for us . up to second order in @xmath129 , the mixed - spin eigenstates are given by [ mixedspin= ] @xmath130 . to the same order , their energies are given by @xmath131 the states with @xmath132 from the @xmath70 valley and those with @xmath133 from the @xmath71 valley are called the bright @xmath4 excitons . those with the opposite @xmath121 have a higher energy ( due to the @xmath134 term ) and are called the bright @xmath5 excitons . their energies are @xmath135 the states obtained by flipping the spin in the conduction band can be called the dark @xmath4 and @xmath5 excitons , their energies are @xcite @xmath136 we calculate the susceptibility @xmath91 from eqs . ( [ chidef= ] ) using the matrix elements @xmath137 and @xmath138 , which , in turn , are obtained from expressions ( [ mixedspin= ] ) for the mixed - spin eigenstates @xmath102 and from the matrix elements between the zero - approximation states ( [ state0= ] ) following from the definitions of the in - plane current and the @xmath36-dipole moment operators [ eqs . ( [ current= ] ) and ( [ hrashba= ] ) ] . namely , for the in - plane current we have @xmath139 the nonzero matrix elements of the @xmath36-dipole moment operator between the ground state and the zero - approximation states are @xmath140 summation over @xmath141 in combination with the independence of the energies @xmath142 on the valley index @xmath111 restores the in - plane isotropy in the final expressions for the susceptibility , written to the second order in @xmath68 : [ chi=]@xmath143 ^ 2}{e_a^2 } \left[\frac{1-w_{\mathbf{q}}}{e_a^\mathrm{b}({\mathbf{q}})-\omega } + \frac{w_{\mathbf{q}}}{e_a^\mathrm{d}({\mathbf{q}})-\omega}\right ] \nonumber \\ & \hspace{1.5cm}{}+2\delta_{\alpha\beta } \frac{[ev\phi(0)]^2}{e_b^2}\left [ \frac{1-w_{\mathbf{q}}}{e_b^\mathrm{b}({\mathbf{q}})-\omega } + \frac{w_{\mathbf{q}}}{e_b^\mathrm{d}({\mathbf{q}})-\omega}\right ] , \label{chinplane=}\\ & \chi_{{z}\alpha}({\mathbf{q}},\omega)=-\chi_{\alpha{z}}({\mathbf{q}},\omega)=\nonumber \\ & \quad=\frac{iq_\alpha\lambda_\mathrm{ex}^}{\delta_c}\ , \frac{ev\phi^2(0)}{e_a}\,d_z \left[\frac{1}{e_a^\mathrm{b}({\mathbf{q}})-\omega } -\frac{1}{e_a^\mathrm{d}({\mathbf{q}})-\omega}\right],\\ & \chi_{zz}({\mathbf{q}},\omega)=2\|d_z\|^2\phi^2(0)\ , \frac{1}{e_a^\mathrm{d}({\mathbf{q}})-\omega}.\end{aligned}\ ] ] here we omitted the infinitesimal imaginary part in the denominator to keep the formulas more compact and approximated @xmath144 in the non - resonant prefactor , neglecting the dispersion and the conduction band splitting . note also that @xmath145 as well as @xmath57 are real quantities . from eqs . ( [ chi= ] ) , we see the crucial role played by the quantity @xmath146 , introduced in sec . [ ssec : excitons ] . it represents the spectral weight transferred from the bright to the dark excitons by the rashba coupling . let us estimate its order of magnitude . taking the numerical values typical of wse@xmath1 , @xmath147 , the experimentally determined splitting @xmath148 @xcite , and @xmath149 from ref . @xcite for a quite strong perpendicular electric field @xmath150 , we still obtain a very small value of @xmath151 . if instead of the perpendicular electric field @xmath64 , an in - plane magnetic field @xmath152 is applied , spin mixing due to the zeeman effect can be taken into account by full analogy with the rashba mixing . in fact , it is sufficient to replace @xmath153 , so eq . ( [ chinplane= ] ) has the same form , but the transferred spectral weight @xmath154 . for a magnetic field @xmath155 , we obtain a numerical estimate @xmath156 . the in - plane direction of the off - diagonal component @xmath157 is determined not by @xmath93 , but by the magnetic field , @xmath158 . here we consider the interaction of excitons in the tmdc monolayer with the electromagnetic field . the monolayer is assumed to be sandwiched between two semi - infinite media with dielectric constants @xmath159 and @xmath160 occupying the half - spaces with @xmath161 and @xmath162 , respectively . ( note that the values of @xmath163 at optical frequencies @xmath164 should be taken . ) the free field in such a structure is fully characterized by the green s function @xmath165 , which represents the response of the electric field , @xmath166 , to an external oscillating polarization @xmath167 , located in the plane @xmath168 . in the quantum theory , this green s function represents the retarded propagator of the electric field ; at the same time , it can be found from the classical maxwell equations @xcite . the radiative self - energy for the excitons at @xmath169 is proportional to @xmath170 , for which we introduce the short - hand notation @xmath171 [ more precisely , it represents the projection on the spatial profile of the excitonic polarization in the @xmath36-direction , here assumed to be just @xmath172 . the long - range exchange part of the coulomb interaction shifts the exciton energies and lifts the valley degeneracy . exciton coupling to the photons also shifts the excitonic frequencies and leads to the radiative decay . all these effects can be described by studying the linear system @xmath173 for each @xmath93 , it has non - trivial solutions for some complex values of @xmath92 whose real parts give the shifted exciton energies , and the imaginary parts ( with the opposite sign and multiplied by 2 ) represent the radiative decay rates . this procedure is equivalent to finding the poles of the full layer susceptibility ( i. e. , dressed by the exchange interaction and coupling to photons ) in the complex plane of @xmath92 , or to finding the poles of the monolayer reflectivity . calculation of @xmath171 from the maxwell equations is quite standard and is given in appendix [ app : maxwell ] . the result is @xmath174 here @xmath175 and @xmath176 are the @xmath36 components of the three - dimensional wave vector in the corresponding media : @xmath177 where the infinitesimal imaginary part determines the behaviour in the complex plane of @xmath92 , prescribing @xmath178 for real @xmath92 in the interval @xmath179 , which corresponds to evanescent waves . the parameter @xmath180 in eq . ( [ dzz= ] ) represents the electric field of a double layer , arising from the excitonic polarization in the @xmath36 direction ( appendix [ app : maxwell ] ) , which is singular in the limit of an infinitely thin layer , and can not be determined in the macroscopic framework , used here . for excitons in semiconductor quantum wells of a sizable width , this local - field effect could be treated properly in the envelope - function approximation @xcite . in the atomically - thin tmdc monolayer , it requires the full microscopic treatment of the short - range exchange interaction , such as that in ref . the consequences of this local - field effect will be discussed in the next subsection . the tensor structure of @xmath91 and @xmath181 , following from the in - plane isotropy of the problem , determines how the two - fold valley degeneracy of the mechanical excitons is lifted . the first family of solutions of eq . ( [ psystem= ] ) is characterized by @xmath94 lying in the @xmath13 plane , perpendicular to @xmath93 . these transverse excitons emit @xmath182-polarized light ( transverse - electric , or te modes ) . the second family of solutions has @xmath94 in the plane formed by the vectors @xmath93 and @xmath183 ; its precise direction is determined by the relative magnitude of different components of @xmath91 . these are longitudinal excitons , which emit @xmath184-polarized light ( transverse - magnetic , or tm modes ) . each of the two linear polarizations represents a linear combination of the two circularly polarized excitons from each valley with equal probability weights . in the following , we treat eq . ( [ psystem= ] ) perturbatively . namely , we assume @xmath92 to be close to one of the poles @xmath103 of @xmath91 . near the pole , it can be represented as @xmath185 with some vectors @xmath186 , @xmath187 , which describe the polarization of the transverse and longitudinal excitons . then , the corresponding solutions of eq . ( [ psystem= ] ) are @xmath188 and @xmath189 , and the shifted ( complex ) excitonic energies are approximately given by @xmath190 here , @xmath191 is nothing but the radiative self - energy of the exciton , in which we explicitly separate the radiative shift @xmath192 and the decay rate @xmath193 . this procedure is equivalent to the quantum - mechanical calculation of the energy shift by perturbation theory to the first order in the coulomb exchange and to the second order in the exciton - photon coupling , and to the calculation of the radiative decay rate by the fermi golden rule . if the spin mixing is due to the zeeman effect in an in - plane magnetic field @xmath84 instead of the rashba effect , the @xmath182 and @xmath184 polarizations do not separate , as @xmath93 and @xmath84 define two different directions in the plane . we discuss this situation in appendix [ app : zeeman ] . for the transverse bright and dark @xmath4 excitons , the radiative shift and decay rate are straightforwardly evaluated as @xmath194 where the upper / lower line in the braces with @xmath146 refers to the bright / dark exciton , and the expression to the left of the braces is the decay rate of the transverse bright @xmath4 exciton in the absence of spin flip @xcite . in the square roots , @xmath195 , we neglected the small difference between @xmath196 and @xmath197 as well as their dependence on @xmath93 , since we are interested only in a narrow region @xmath198 where at least one of @xmath175 , @xmath176 is real @xcite . for the longitudinal @xmath4 excitons we obtain more bulky expressions , @xmath199,\label{gammalab=}\\ \omega_{la}^\mathrm{d}({\mathbf{q } } ) -\frac{i}2\,\gamma_{la}^\mathrm{d}({\mathbf{q } } ) = { } & { } -\frac{8\pi{i}q^2\phi^2(0)}{q_{1z}^a/{\varepsilon}_1+q_{2z}^a/{\varepsilon}_2 } \left[\frac{e^2v^2}{e_a^2 } \frac{\|\lambda_\mathrm{ex}\|^2}{4\delta_\mathrm{c}^2}\ , \frac{q_{1z}^aq_{2z}^a}{{\varepsilon}_1{\varepsilon}_2}+\|d_z\|^2\right]+{}\nonumber\\ { } & { } + \frac{8\pi{}q^2\phi^2(0)}{q_{1z}^a/{\varepsilon}_1+q_{2z}^a/{\varepsilon}_2}\ , \frac{ev}{e_a}\frac{\lambda_\mathrm{ex}^}{2\delta_\mathrm{c } } \left(\frac{q_{1z}^a}{{\varepsilon}_1}-\frac{q_{2z}^a}{{\varepsilon}_2}\right)d_z + 4\pi\kappa_0\|d_z\|^2\phi^2(0).\label{gammalad=}\end{aligned}\ ] ] whose qualitative features are similar to the transverse case : the decay rates are nonzero only in the small - momentum region @xmath198 , the bright exciton radiative rate is dominated by the first line of eq . ( [ gammalab= ] ) to which the rashba term gives a small correction . however , an important difference from the transverse case is that @xmath200 does not vanish for @xmath201 . the longitudinal dark @xmath4 exciton can decay even in the absence of the rashba coupling , due to @xmath68 . moreover , this latter mechanism is by far the dominant one , as the ratio between the two terms in the square brackets in the first line of eq . ( [ gammalad= ] ) is @xmath202 for a quite high electric field of @xmath150 . the last term in eq . ( [ gammalad= ] ) containing the unknown parameter @xmath180 is purely real and does not contribute to the radiative decay rate . however , it produces an energy shift , @xmath203 of the longitudinal excitons , which thus lifts the valley degeneracy even at @xmath204 and in the absence of rashba or zeeman effects @xcite . in the macroscopic framework , used here , @xmath205 represents the interaction energy of the exciton polarization in the @xmath36 direction with its own electric field , which is singular in the limit of an infinitely thin layer . by the order of magnitude , @xmath206 , where @xmath3 is the monolayer thickness and @xmath207 is the effective background dielectric constant in the perpendicular direction at optical frequencies ( see appendix [ app : maxwell ] ) . taking @xmath208 , @xmath209 , @xmath210 , and @xmath211 , we obtain an estimate @xmath212 . this value agrees by the order of magnitude with the dark exciton energy shifts due to the short - range exchange interaction , which were calculated using a microscopic _ ab initio _ approach in ref . indeed , both are supposed to have the same origin , as the coulomb part of @xmath171 represents the exchange field . however , no lifting of valley degeneracy was mentioned in ref . note that the presence of the positive shift @xmath205 for the longitudinal excitons has an important consequence for the luminescence , as the lowest - energy states are the transverse excitons whose decay rate is only due to rashba or zeeman effect . to illustrate the @xmath128 dependence of the radiative energy shifts and the decay rates , we plot them in figs . [ fig : ta ] and [ fig : la ] for the transverse and longitudinal dark @xmath4 excitons , for @xmath213 and two values of the substrate dielectric constant , one typical for glass - like substrates , @xmath214 , the other one corresponding to a highly dielectric substrate , such as algasb with @xmath215 at optical frequencies @xcite . it is convenient to normalize the energy shifts and decay rates by the bright exciton decay rate at @xmath204 , @xmath216 , for a tmdc monolayer suspended in vacuum , @xmath217 for @xmath218 , @xmath219 , @xmath220 , this estimate gives @xmath221 , not very much different from the results of more precise calculations involving the full microscopic treatment of the exciton wave function @xcite , which give @xmath222 for several tmdc materials . in the calculation presented in figs . [ fig : ta],[fig : la ] , we used the values @xmath219 , the experimentally determined splitting @xmath148 @xcite , @xmath210 as extracted from ref . @xcite , and the rashba coupling constant @xmath223 per each v / of the static electric field @xmath64 @xcite . the latter was taken to be 0.1 v / for the glass - like substrate , giving @xmath149 , and 0.2 ev / for the highly dielectric substrate ( @xmath224 ) , as estimated in appendix [ app : substrate ] ( note that changing the electric field amounts to a simple rescaling of the @xmath225 axis @xmath226 in fig . [ fig : ta ] , while fig . [ fig : la ] is insensitive to @xmath64 , as discussed above ) . ( black curves , negative @xmath225 axis ) and decay rates @xmath227 ( red curves , positive @xmath225 axis ) of transverse excitons for @xmath228 , @xmath229 ( solid curves ) and @xmath230 ( dashed curves ) , as a function of the dimensionless exciton momentum @xmath231.,width=321 ] ) @xmath232/\gamma_0 $ ] ( black curves , negative @xmath225 axis ) and decay rates @xmath233 ( red curves , positive @xmath225 axis ) of longitudinal excitons for @xmath228 , @xmath229 ( solid curves ) and @xmath230 ( dashed curves ) , as a function of the dimensionless exciton momentum @xmath231 . , width=321 ] so far , we calculated the decay rates for excitons with a given momentum @xmath93 . if we want to characterize the radiative decay of a thermal population of excitons , these rates should be averaged over their thermal distribution @xcite . this assumes that the thermalization rate is larger than the radiative rate , otherwise the full kinetic problem should be analyzed . we assume the exciton distribution to be non - degenerate : @xmath234 where @xmath235 is the combined density of the @xmath4 excitons , both bright and dark ones , @xmath236 is the temperature , and we introduced the shorthand notation for the exciton mass , @xmath237 . the distribution functions @xmath238 are spread over the thermal momenta @xmath239 . then , the thermal average can be approximated as @xmath240 where the summation is over the four species of @xmath4 excitons . taking the leading contributions to expressions ( [ gammata= ] ) , ( [ gammalab= ] ) , ( [ gammalad= ] ) , we perform the integration analytically ( see appendix [ app : integral ] ) , and obtain @xmath241 the explicit form of the functions @xmath242 , @xmath243 , @xmath244 is given in appendix [ app : integral ] . for @xmath213 , @xmath214 , they are equal to 1.5 , 1.6 , and 1.0 , respectively . much more important are the factors @xmath245 , @xmath246 , which characterize the relative oscillator strength of the dark @xmath4 excitons due to the rashba - induced in - plane dipole moment and due to the perpendicular dipole moment . they are defined as @xmath247 and their values are estimated to be @xmath248 for the field @xmath150 , and @xmath249 . if instead of a perpendicular electric field , an in - plane magnetic field is applied , we use the results of appendix [ app : zeeman ] . the result amounts to the replacement @xmath250 in the first and the second terms in the numerator of eq . ( [ avgamma= ] ) , so instead of the rashba factor @xmath245 we have the zeeman factor @xmath251 for a field @xmath155 and @xmath252 , it is about @xmath253 . in this work we have investigated the radiative decay of spin - forbidden dark intravalley @xmath4 excitons in tungsten dichalcogenide monolayers due to several spin - flip mechanisms . our main result is represented by eq . ( [ avgamma= ] ) for the thermally - averaged radiative decay rate and by expressions ( [ etar=])([etaz= ] ) for the average efficiencies of different spin - flip mechanisms . as the bright excitons in tungsten - based compounds have higher energy than the dark ones , the bright exciton contribution to the average rate is suppressed at low temperatures by the thermal activation exponential @xmath254 with the bright - dark splitting @xmath255 of several tens of mev , of the order of the room temperature . the intrinsic mechanism of the dark exciton radiative decay is due to the interband spin - flip dipole moment , perpendicular to the monolayer plane . it was mentioned in refs . @xcite , and its magnitude can be deduced from the estimates of ref . we found , however , that the very same mechanism lifts the valley degeneracy of the dark excitons by a coulomb local - field effect , producing a splitting @xmath205 which we could very roughly estimate as about 10 mev . crucially , the whole oscillator strength of such interband spin flip ( about @xmath256 of the bright exciton oscillator strength ) is taken by the higher - energy component . thus , the spin - forbidden intravalley exciton , usually referred to as dark , in fact , has two components , one dark and the other which can be called `` dim '' . the contribution of this dim exciton to the average radiative rate ( [ avgamma= ] ) is suppressed at low temperatures by another exponential , @xmath257 . for the lowest - energy dark component , which should dominate the exciton population at low temperatures , we have analyzed extrinsic spin - flip mechanisms , which transfer some oscillator strength from the bright exciton to the dark one . one such mechanism is due to the rashba spin - orbit coupling which arises whenever the reflection symmetry in the monolayer plane is broken , e. g. , by an externally applied perpendicular static electric field , or by different dielectric environment above and below the monolayer . this mechanism was mentioned in ref . @xcite . using our own estimate of @xmath258 for the effective electric field produced by the dielectric mismatch and the calculated value of the rashba coupling constant from ref . @xcite , we find that the fraction of the oscillator strength transferred from the bright exciton to the dark one is extremely low , about @xmath6 , mainly due to small momenta of excitons subject to radiative decay . another possible mechanism is the spin flip by the zeeman effect from a magnetic field directed along the monolayer plane . then , for a strong but still realistic field of 30 t , more than @xmath259 of the bright exciton oscillator strength can be transferred to the dark one . this suggests a way to manipulate the radiative properties of dark excitons . the results of the present work help to identify several directions for improvement of our understanding of the excitonic radiative processes in tungsten dichalcogenides . first , the value of the interband spin - flip dipole moment is currently known from only one source , the estimate of ref . . it would be quite helpful if more information were available , either theoretically or experimentally . e. g. , by studying the effect of the rashba spin - orbit coupling on the carrier transport , as the rashba coupling has the same origin . second , a reliable microscopic calculation of the dark - dim exciton splitting @xmath205 would help to determine how dramatic is the low - temperature suppression of the dim exciton population . finally , the role of collisions with defects , charge carriers , or phonons in the radiative decay also needs clarification . we are grateful to m. potemski for stimulating discussions . a. o. s. acknowledges financial support from the ec graphene flagship project ( no . the @xmath260 matrix @xmath261 , can be transformed to the block - diagonal intra - band form by applying a unitary transformation @xmath262 ( here @xmath64 is only the static electric field , while the same term with the optical field remains inter - band ) : @xmath263 e^{\mathcal{s}_\tau({\mathbf{k}})}\ ] ] if the matrix @xmath264 is chosen in the form @xmath265,\\ & \mathcal{s}_{-1}({\mathbf{k}})=\frac{1}{e_\mathrm{g}}\left[\begin{array}{cccc } 0 & 0 & vk_+ & \mathcal{e}_zd_z^ \\ 0 & 0 & 0 & vk_+ \\ -vk_- & 0 & 0 & 0 \\ -\mathcal{e}_zd_z & -vk_- & 0 & 0 \end{array}\right],\end{aligned}\ ] ] where @xmath266 , then , to the second order in @xmath267 , @xmath268 , we obtain a block - diagonal hamiltonian . the first term represents the kinetic energy of the electrons and holes , @xmath269,\ ] ] while the second one corresponds to the rashba coupling , @xmath270.\ ] ] here we also assumed @xmath271 . tmdc materials can have comparable values of the spin splitting in the valence band @xmath134 and the gap @xmath25 ; then the above calculation can be repeated without assuming @xmath271 , which would produce different rashba coupling in the conduction and the valence band . for tungsten compounds @xmath272 , so we neglect this difference . we also neglect the effect of the unitary transformation @xmath273 on the coulomb interaction which forms the excitonic states . an external electric field @xmath64 and interaction with a substrate via van der waals forces have the same qualitative effect of breaking the reflection symmetry in the @xmath36 direction and deforming the electronic wave functions in the tmdc monolayer . still , there is no reason for the deformation to be quantitatively similar in the two cases . thus , characterization of the substrate effect by a single parameter , an effective @xmath64 , is quite a rough approximation , which gives only an order - of - magnitude estimate of the effect . furthermore , we adopt a macroscopic description of the substrate , characterizing it by the dielectric constant . as the substrate is only a few angstroms away from the monolayer center , such macroscopic description is also valid only qualitatively . we model the tmdc monolayer as a slab of thickness @xmath3 and dielectric constant @xmath274 , sandwiched between two media with dielectric constants @xmath159 at @xmath275 and @xmath160 at @xmath276 . let us write the energy as a functional of the microscopic three - dimensional electronic density @xmath277 , @xmath278 where the substrate effect is included as a hartree - like term : @xmath279={}&{}e_0[\rho({\mathbf{r}})]-{}\nonumber\\ { } & { } -\int\left[v({\mathbf{r}},{\mathbf{r}}')-v_1({\mathbf{r}},{\mathbf{r}}')\right ] \times\nonumber\\ { } & \qquad{}\times \rho({\mathbf{r}})\,\rho_\mathrm{i}({\mathbf{r}}')\,d^3{\mathbf{r}}\,d^3{\mathbf{r}}'+\nonumber\\ { } & { } + \frac{1}{2}\int \left[v({\mathbf{r}},{\mathbf{r}}')-v_1({\mathbf{r}},{\mathbf{r}}')\right ] \times\nonumber\\ { } & \qquad{}\times \rho({\mathbf{r}})\,\rho({\mathbf{r}}')\,d^3{\mathbf{r}}\,d^3{\mathbf{r } } ' . \label{hartree=}\end{aligned}\ ] ] here @xmath280 $ ] is the functional for a tmdc monolayer suspended in vacuum , @xmath281 is the ionic density ( a sum of @xmath282 functions at the ion positions ) , @xmath283 is the interaction potential , represented by the green s function of the poisson equation in the full dielectric structure multiplied by @xmath284 , @xmath285 is the same for @xmath286 . the second and the third terms in eq . ( [ hartree= ] ) describe interaction of the tmdc electrons with the polarization charges in the surrounding media . the deformed electronic wave functions can be found by varying the functional ( [ hartree= ] ) . the whole description is analogous to the macroscopic description of polarons in ionic crystals @xcite ; indeed , deformation of electronic wave functions in the tmdc by interaction with a dielectric substrate can be viewed as a polaronic effect . in the planar geometry considered here , evaluation of the integrals in eq . ( [ hartree= ] ) reduces to a summation over image charges . when all three dielectric constants @xmath287 are different , each charge produces an infinite number of images : @xmath288 ^ 2}},\\ & \zeta_0=1,\quad \zeta_{\pm{1}}=\frac{{\varepsilon}-{\varepsilon}_{1,2}}{{\varepsilon}+{\varepsilon}_{1,2}},\quad \zeta_{\pm{2}}=\zeta_{\pm{1}}\ , \frac{{\varepsilon}-{\varepsilon}_{2,1}}{{\varepsilon}+{\varepsilon}_{2,1}},\quad\ldots.\end{aligned}\ ] ] for @xmath289 , the decay of @xmath290 with @xmath291 is quite slow ( which is a manifestation of the confinement of the electric field lines to the interior of the dielectric ) , so many images contribute to the hartree terms in eq . ( [ hartree= ] ) . however , if we start with @xmath277 , symmetric with respect to @xmath292 , the image charge distribution due to @xmath293 is also symmetric and does not produce any net electric field in the @xmath36 direction . as concerns the @xmath294 term , if at least one of @xmath295 , @xmath296 is not small compared to @xmath297 , which we assume to be the case , @xmath290 decays exponentially with a decrement @xmath298 . thus , for an order - of - magnitude estimate , we can restrict ourselves to @xmath299 . the electronic states we are interested in ( the bottom of the conduction band and the top of the balence band at @xmath2 points ) are known to originate mainly from @xmath3 orbitals of the metal atoms . assuming these orbitals to be concentrated around @xmath169 , we can expand the potentials in @xmath300 as a formal small parameter . of course , in reality the orbital spatial extent is of the same order as @xmath3 , so this multipole expansion is good only for an order - of - magnitude estimate . the overall charge neutrality and the vanishing dipole moment of the unperturbed charge distribution around each metal atom ( due to its @xmath301 symmetry ) make the first non - vanishing multipolar moment to be the quadrupole . the quadrupole field decays with distance as @xmath302 , so , even though each atom feels the field from images of all other atoms , the dominant contribution comes from the nearest two images which are due to the atom itself . the electric field at the point @xmath303 produced by the two quadrupole images at @xmath304 is given by @xmath305 where @xmath306 is the quadrupole moment of the electronic cloud around a single metal atom , @xmath307 although the quadrupole moment of a metal atom deformed by the crystal field differs from that of an isolated atom , we assume them to be of the same order and estimate the latter . the electronic configuration of tungsten is [ xe]@xmath308 , and four out of five outermost @xmath3 orbital states are filled , according to the hund s rule . since filled atomic shells have zero quadrupole moment , @xmath306 for tungsten is determined by the single empty @xmath309 orbital which we take to be the @xmath310 one , as it is this one that is known to give rise to the lowest conduction band of tungsten dichalcogenides . the wave function of this orbital can be approximated by a hydrogen - like one with @xmath311 , which moves in the coulomb potential with effective charge @xmath312 the average quadrupole moment of a hydrogenic @xmath313 state @xcite , @xmath314}{2}\,\frac{2l(l+1)-6m^2}{4l(l+1)-3},\ ] ] then gives @xmath315 for tungsten ( @xmath316 is the bohr radius ) . for molybdenum , the electronic configuration in the gaseous phase , [ kr]@xmath317 , has a filled @xmath318 shell with zero quadrupole moment . however , given the fact that the lowest conduction band of molybdenum dichalcogenides is formed mostly by the @xmath318 @xmath310 orbital , just like for tungsten dichalcogenides , the electronic configuration is likely to be changed by the crystal field to [ kr]@xmath319 . then , repeating the same calculation with @xmath320 , @xmath321 @xcite , we obtain @xmath322 for molybdenum . if a tmdc monolayer suspended in vacuum is subjet to an _ external _ electric field @xmath64 , the field inside the monolayer is @xmath323 . it is this one that should be matched with @xmath324 from eq . ( [ ez0= ] ) to obtain the effective external field due to the substrate : @xmath325 taking the static dielectric constants @xmath326 ( see ref . @xcite and references therein ) , @xmath213 ( vacuum ) , @xmath327 ( silica ) , and @xmath328 , we obtain @xmath329 for tungsten dichalcogenides . taking a highly dielectric substrate , @xmath330 , one can increase the effective field by about a factor of 2 . here we solve the maxwell equations in the presence of the oscillating monolayer polarization @xmath331 sandwiched between two semi - infinite media with dielectric functions @xmath159 and @xmath160 occupying the half - spaces with @xmath161 and @xmath162 , respectively . due to the in - plane isotropy of the maxwell equations and of the susceptibility , @xmath332 , @xmath333 , we can assume the wave vector @xmath93 to be along the @xmath334 axis , without any loss of generality . for the three - dimensional waves propagating in the two media , the @xmath36 component of the wave vector is given by @xmath335 . for the @xmath182 polarization , the fields in the incident , reflected and transmitted waves are parametrized by their electric field amplitudes @xmath336 , @xmath337 and @xmath338 , respectively . the non - zero components of the fields are given by ( we omit the common factor @xmath339 ) @xmath340 the boundary conditions on the tangential components of the electric and magnetic fields correspond to the continuity of the electric field and a jump in the magnetic field due to the surface current : @xmath341 when the incident wave is absent , but the layer polarization acts as a source producing the outgoing field , the amplitudes of the latter are given by @xmath342 for the @xmath184 polarization , we choose the magnetic field amplitudes @xmath343 , @xmath344 , @xmath345 , to parametrize the fields , which are sought in the form [ tm=]@xmath346 the boundary conditions in this case are more subtle , because the @xmath36-polarization @xmath347 produces an electrical double layer . let us assume @xmath348 to be spread over a narrow but finite region @xmath349 ( @xmath3 being the monolayer thickness ) with a background dielectric constant @xmath297 . in the first maxwell equation for @xmath350 , @xmath351,\ ] ] the first term is finite at @xmath352 . in the layer region it can be neglected , which gives the following result for the field in the layer region ( we omit the terms which vanish at @xmath352 ) : @xmath353 where the first term represents the field at @xmath354 , related to the value at @xmath355 from eq . ( [ tmez= ] ) by the continuity of the normal component of the electric displacement . in the third maxwell equation for @xmath356 , the @xmath348 term appears both in the polarization and displacement currents , so it cancels out , and the only singularity in @xmath357 comes from the in - plane current , @xmath358 however , from the faraday s law , @xmath359 it follows that @xmath360 must have a jump of @xmath361 : @xmath362 this leads to the following boundary conditions for the amplitudes in eqs . ( [ tm= ] ) : @xmath363 if we wish to determine the outgoing field produced by a source layer polarization without the incident field , we have no ambiguity ; the amplitudes are given by @xmath364 however , we face a problem when we want to couple the field back to the polarization , as we need the value @xmath365 , which is undetermined due to the singularity . here we note that coupling to the layer polarization is in fact determined by @xmath366 , with the same @xmath348 as in the spatial profile of the polarization itself , which is nothing but the product of the microscopic wave functions of the electron and the hole at coinciding points ( see , e. g. , a microscopic treatment for excitons in a semiconductor quantum well @xcite ) . in other words , we define @xmath367 i. e. , as the projection on the polarization spatial profile in the @xmath36 direction . then , the uncertainty due to the integral terms in eqs . ( [ ez = delta ] ) , ( [ ex = delta ] ) is resolved as @xmath368 by symmetry . on the contrary , the term @xmath369 can only be determined from the microscopic theory . the projected fields are given by @xmath370 to arrive at the final expressions ( [ dab=])([dzz= ] ) , one should recall that the field @xmath64 which appears in the definition of the dipole moment ( [ hrashba= ] ) , and which was used in the rashba coupling constant estimate of ref . @xcite , is not the local field , but the external field , applied to a monolayer suspended in vacuum . hence , a factor @xmath371 should be absorbed into @xmath68 . as the @xmath182 and @xmath184 polarizations do not separate in a parallel magnetic field , we return to the valley basis . namely , instead of using eq . ( [ chipole= ] ) , we represent the susceptibility near each pole @xmath103 as @xmath372 where the vector @xmath373 , describes the polarization of the exciton @xmath104 in the valley @xmath111 . then , the radiative self - energy can be approximated by @xmath374 for the dark @xmath4 excitons in the parallel magnetic field , these vectors are given by @xmath375 where @xmath376 . as we have seen in sec . [ ssec : gammaq ] , the most important term that lifts the valley splitting is the last term in eq . ( [ dzz= ] ) . thus , we first consider the splitting at @xmath204 , @xmath377 , determined by the self - energy matrix @xmath378,\ ] ] where @xmath379 is the phase of @xmath68 . its eigenvectors determine the @xmath36-dipole - active state @xmath380 whose energy is shifted up by @xmath205 , and the orthogonal one , @xmath381 , which is @xmath36-dipole - inactive . in the case of rashba - induced splitting at @xmath382 these states evolve into the @xmath184- and @xmath182-polarized ones , respectively . at finite @xmath93 and @xmath84 , we project the self - energy on the two eigenvectors @xmath383 , which gives the radiative shift and the decay rate of the two states @xmath384\nonumber\\ & + 4\pi\phi^2(0)\,\frac{ev\|d_z\|}{e_a}\ , \frac{g_\\|\mu_bb_\\|}{4\delta_\mathrm{c}}\,\frac{q_{1z}^a/{\varepsilon}_1-q_{2z}^a/{\varepsilon}_2}{q_{1z}^a/{\varepsilon}_1+q_{2z}^a/{\varepsilon}_2}\,q\cos\tilde\phi -\frac{8\pi{i}q^2\|d_z\|^2\phi^2(0)}{q_{1z}^a/{\varepsilon}_1+q_{2z}^a/{\varepsilon}_2}+\xi_0,\\ \omega_\\|({\mathbf{q}})-\frac{i}2\,\gamma_\\|({\mathbf{q}})= { } & { } -8\pi{i}\,\phi^2(0)\,\frac{e^2v^2}{e_a^2 } \left(\frac{g_\\|\mu_bb_\\|}{4\delta_\mathrm{c}}\right)^2 \left[\frac{(e_a / c)^2}{q_{1z}^a+q_{2z}^a}\cos^2\tilde\phi + \frac{(q_{1z}^a/{\varepsilon}_1)(q_{2z}^a/{\varepsilon}_2)}{q_{1z}^a/{\varepsilon}_1+q_{2z}^a/{\varepsilon}_2}\sin^2\tilde\phi\right ] , \label{gammab=}\end{aligned}\ ] ] where @xmath385 , and @xmath386 are the polar angles of @xmath93 and @xmath84 in the @xmath13 plane , respectively . it is convenient to introduce a dimensionless variable @xmath387 , then the integration over @xmath93 transforms as @xmath388 , and the radiative decay rates of the @xmath4 excitons , given by eqs . ( [ gammata= ] ) , ( [ gammalab= ] ) , ( [ gammalad= ] ) , ( [ gammab= ] ) , can be written as where we keep only the dominant terms . it is convenient to get rid of the square roots in the denominators and take the real part of each term in the numerators separately . then the integration becomes straightforward and gives @xmath390 , \label{fbta=}\\ \int\frac{d^2{\mathbf{q}}}{(2\pi)^2}\ , \frac{\gamma^\mathrm{d}_{la}({\mathbf{q}})}{\gamma_0}= { } & { } \eta_z\,\frac{(e_a / c)^2}{2\pi } \left[\frac{2{\varepsilon}_1 ^ 2{\varepsilon}_2 ^ 2}{({\varepsilon}_1+{\varepsilon}_2)^2}\ , \frac{\sqrt{{\varepsilon}_1{\varepsilon}_2}+({\varepsilon}_1+{\varepsilon}_2)/3}{\sqrt{{\varepsilon}_1}+\sqrt{{\varepsilon}_2 } } \right.+\nonumber\\ & \hspace{3cm}{}+\left . \frac{2{\varepsilon}_1 ^ 3{\varepsilon}_2 ^ 3}{({\varepsilon}_2 ^ 2-{\varepsilon}_1 ^ 2)({\varepsilon}_1+{\varepsilon}_2)^{3/2 } } \ln\left(\sqrt{\frac{{\varepsilon}_2}{{\varepsilon}_1}}\ , \frac{\sqrt{{\varepsilon}_2+{\varepsilon}_1}+\sqrt{{\varepsilon}_2 } } { \sqrt{{\varepsilon}_2+{\varepsilon}_1}+\sqrt{{\varepsilon}_1}}\right)\right ] . \label{fdta=}\end{aligned}\ ] ] k .- k . liu , w. zhang , y .- h . lee , y .- c . chang , c .- y . su , c .- s . chang , h. li , y. shi , h. zhang , c .- s . lai , and l .- j . li , _ growth of large - area and highly crystalline mos@xmath1 thin layers on insulating substrates _ , nano . lett . 12 * , 1538 ( 2012 ) . y. zhang , t .- chang , b. zhou , y .- t . cui , h. yan , z. liu , f. schmitt , j. lee , r. moore , y. chen , h. lin , h .- t . jeng , s .- k . mo , z. hussain , a. bansil , and z .- x . shen , _ direct observation of the transition from indirect to direct bandgap in atomically thin epitaxial mose@xmath1 _ , nature nanotech . * 9 * , 111 ( 2014 ) . a. chernikov , t. c. berkelbach , h. m. hill , a. rigosi , y. li , o. b. aslan , d. r. reichman , m. s. hybertsen , and t. f. heinz , _ exciton binding energy and nonhydrogenic rydberg series in monolayer ws@xmath1 _ , phys . lett . * 113 * , 076802 ( 2014 ) . g. sallen , l. bouet , x. marie , g. wang , c. r. zhu , w. p. han , y. lu , p. h. tan , t. amand , b. l. liu , and b. urbaszek , _ robust optical emission polarization in mos@xmath1 monolayers through selective valley excitation _ , phys . rev . b * 86 * , 081301(r ) ( 2012 ) . g. wang , l. bouet , d. lagarde , m. vidal , a. balocchi , t. amand , x. marie , and b. urbaszek , _ valley dynamics probed through charged and neutral exciton emission in monolayer wse@xmath1 _ , phys . b * 90 * , 075413 ( 2014 ) . k. hao , g. moody , f. wu , c. k. dass , l. xu , c .- h . chen , l. sun , m .- y . li , l .- j . li , a. h. macdonald , and x. li , _ direct measurement of exciton valley coherence in monolayer wse@xmath1 _ , nature phys . ( 2016 ) . a. kormnyos , v. zlyomi , n. d. drummond , p. rakyta , g. burkard , and v. i. falko , _ monolayer mos@xmath1 : trigonal warping , the @xmath395 valley , and spin - orbit coupling effects _ , phys . b * 88 * , 045416 ( 2013 ) . f. withers , o. del pozo - zamudio , s. schwarz , s. dufferwiel , p. m. walker , t. godde , a. p. rooney , a. gholinia , c. r. woods , p. blake , s. j. haigh , k. watanabe , t. taniguchi , i. l. aleiner , a. k. geim , v. i. falko , a. i. tartakovskii , and k. s. novoselov , _ wse@xmath1 light - emitting tunneling transistors with enhanced brightness at room temperature _ , nano lett . * 15 * , 8223 ( 2015 ) . g. wang , c. robert , a. suslu , b. chen , s. yang , s. alamdari , i. c. gerber , t. amand , x. marie , s. tongay , and b. urbaszek , _ spin - orbit engineering in transition metal dichalcogenide alloy monolayers _ , nature comm . * 6 * , 10110 ( 2015 ) . a. a. mitioglu , p. plochocka , . granados del aguila , p. c. m. christianen , g. deligeorgis , s. anghel , l. kulyuk , and d. k. maude , _ optical investigation of monolayer and bulk tungsten diselenide ( wse@xmath1 ) in high magnetic fields _ , nano lett . * 15 * , 4387 ( 2015 ) . m. koperski , k. nogajewski , a. arora , v. cherkez , p. mallet , j .- y . veuillen , j. marcus , p. kossacki , and m. potemski , _ single photon emitters in exfoliated wse@xmath1 structures _ , nature nanotech . * 10 * , 503 ( 2015 ) . t. smoleski , m. goryca , m. koperski , c. faugeras , t. kazimierczuk , k. nogajewski , p. kossacki , and m. potemski , _ tuning valley polarization in a wse@xmath1 monolayer with a tiny magnetic field _ , arxiv:1512.00839 . v. m. agranovich and o. a. dubovskii , _ effect of retarded interaction on the exciton spectrum in one - dimensional and two - dimensional crystals _ , pisma zhetf * 3 * , 345 ( 1966 ) [ jetp lett . * 3 * , 223 ( 1966 ) ] . m. m. glazov , t. amand , x. marie , d. lagarde , l. bouet , and b. urbaszek , _ exciton fine structure and spin decoherence in monolayers of transition metal dichalcogenides _ , b * 89 * , 201302(r ) ( 2014 ) . a. kormnyos , g. burkard , m. gmitra , j. fabian , v. zlyomi , n. d. drummond and v. falko , _ k@xmath396p theory for two - dimensional transition metal dichalcogenide semiconductors _ , 2d mater . * 2 * , 022001 ( 2015 ) . a. v. stier , k. m. mccreary , b. t. jonker , j. kono , and s. a. crooker _ exciton diamagnetic shifts and valley zeeman effects in monolayer @xmath397 and @xmath393 to 65 tesla _ , nat . comm . * 7 * , 10643 ( 2016 ) . one might try to improve the precision of eq . ( [ lambdaex= ] ) by replacing @xmath398 in the denominator which represents the interband excitation energy ; indeed , the excion binding energy @xmath53 is known to be quite large in tmdcs , values up to @xmath399 having been reported @xcite . however , if one corrects the denominator in eq . ( [ lambdaex= ] ) , then , to be consistent , one would also have to correct that in eq . ( [ meh= ] ) , as well as to go beyond the parabolic approximation for the electron and hole dispersion . thus , to be systematic , we assume to be in the limit @xmath400 and neglect corrections of the order of @xmath401 and higher . beyond this limit ( the extreme case being that of frenkel excitons ) , one can still work with the effective excitonic hamiltonian ( [ rashbaex= ] ) , however , its parameters have no simple relation with the single - electron band structure parameters . in fact , the energy difference between the dark and bright excitons does not necessarily coincide with the single - particle spin splitting in the conduction band , due to an extra contribution from the exchange interaction @xcite . here , @xmath255 should be understood as the excitonic splitting . in fact , lifting of the valley degeneracy for the dark @xmath4 excitons follows from a very simple and general consideration . consider the states at @xmath204 in the absence of external fields . then , we have two degenerate excitations corresponding to the @xmath2 valleys , whose transition dipole moments are both directed along @xmath36 . then , it is always possible to form two linear combinations of them , one of which will have zero dipole moment and zero decay rate , while for the other one they will be finite . if the decay rates for the two states are different , so must be the energy shifts , which means lifting of the degeneracy . note that for @xmath402 , the denominator in eq . ( [ gammata= ] ) can vanish ; in this case , the perturbative approach ( [ goldenrule= ] ) is not valid , so a more precise analysis of the original dispersion equation ( [ psystem= ] ) is required @xcite . here we assume @xmath295 and @xmath296 to differ significantly , so this problem does not arise .	we perform a theoretical study of radiative decay of dark intravalley excitons in transition metal dichalcogenide monolayers . this decay necessarily involves an electronic spin flip . the intrinsic decay mechanism due to interband spin - flip dipole moment perpendicular to the monolayer plane , gives a rate about 1001000 times smaller than that of bright excitons . however , we find that this mechanism also introduces an energy splitting due to a local field effect , and the whole oscillator strength is contained in the higher - energy component , while the lowest - energy state remains dark and needs an extrinsic spin - flip mechanism for the decay . rashba effect due to a perpendicular electric field or a dielectric substrate , gives a negligible radiative decay rate ( about @xmath0 times slower than that of bright excitons ) . spin flip due to zeeman effect in a sufficiently strong in - plane magnetic field can give a decay rate comparable to that due to the intrinsic interband spin - flip dipole .
ultraluminous x - ray sources ( ulxs ) are variable x - ray sources which are not coincident with the nucleus of the galaxy and exhibit luminosities over the eddington limit of a 20 @xmath2 black hole ( @xmath3 ) assuming isotropic emission . they are too bright to be powered by accretion onto stellar mass black holes like galactic black hole binaries ( bhbs ) and are thus candidates of intermediate mass black holes ( imbhs ) . however , if the radiation is super - eddington or beamed along the line of sight , imbhs are not required @xcite . modeling the x - ray spectra of ulxs could shed light on their masses . for example , the temperature and size of the accretion disk can be used to weigh the central compact object @xcite . however , identification and quantification of the disk emission in the energy spectrum , especially when it is the dominant component , has confronted difficulties and been proved unreliable in a few cases @xcite . in contrast , characteristic time scales of the x - ray emission , such as quasi - periodic oscillations ( qpos ) and the frequency break in the power spectrum , could be used to determine the mass of compact objects via a model independent calibration ( cf . * ? ? ? * ; * ? ? ? low frequency qpos from mhz to several tens hz have been found in galactic bhbs . on the basis of their properties like the coherence , amplitude , phase lag , and harmonic component , low frequency qpos can be classified into three types of a , b and c @xcite . type a / b qpos appear in a relatively narrow frequency range at a few hz , while type c qpos vary in a wide frequency range in response to the spectral parameters of the source . for example , the type c qpos in grs 1915 + 105 are detected at frequencies of as low as 1 mhz @xcite . qpos have been detected in two ulxs , and are argued to be of type c @xcite because they appear to vary in a manner similar to type c qpos in galactic bhbs . scaling the compact object mass with the oscillation frequency indicates the presence of imbhs in these two ulxs @xcite . m82 is a starburst galaxy at a distance of 3.63 mpc @xcite . the x - ray source was identified as a transient ulx in m82 from multiple x - ray observations with chandra @xcite . @xcite argued that is more likely to be an intermediate mass black hole than a stellar mass object accreting from a massive star according to its transient nature @xcite . it lies on the sky plane at @xmath4 to another ulx , which at most times is the brightest source in m82 , and thus chandra is the only telescope able to spatially resolve them in x - rays . @xcite reported significant timing noise near 1 mhz from from one chandra observation . here , using new chandra data , we report the discovery of mhz qpos in and discuss their possible nature . we checked all archival chandra and xmm - newton observations of m82 to date , as well as three joint chandra / xmm observations that we performed recently ( pi : h. feng ) , to search for timing noise of at low frequencies . among these observations , qpo features at a few mhz are found in five observations ( chandra obsids 6097/8190/10027 and xmm obsids 0112290201/0560590101 ) , two out of which , 10027 and 0560590101 , were made simultaneously with an over lap of 17 ks . lccccc[b ] observation i d & 0112290201 & 6097 & 8190 & 10027 & 0560590101 + observation date & 2001 - 05 - 06 & 2005 - 02 - 04 & 2007 - 06 - 02 & 2008 - 10 - 04 & 2008 - 10 - 03 + exposure ( ks ) & 27 & 53 & 53 & 19 & 30 + @xmath5 ( counts s@xmath6 ) & 2.59 & 0.20 & 0.31 & 0.29 & 3.27 + @xmath7 & 0.7 & 1.0 & 1.0 & 1.0 & 0.2 + + @xmath8 ( @xmath9 @xmath10 ) & & @xmath11 & @xmath12 & @xmath13 & + @xmath14 & & @xmath15 & @xmath16 & @xmath17 & + @xmath18 ( @xmath19 ) & & @xmath20 & @xmath21 & @xmath22 & + @xmath23 ( @xmath0 ) & @xmath24 & @xmath25 & @xmath26 & @xmath27 & + @xmath28 & & 80.7/74 & 87.1/74 & 83.6/74 & + + significance ( @xmath29 ) & 3.00 & 3.39 & 6.51 & 4.49 & 2.91 + frequency ( mhz ) & @xmath30 & @xmath31 & @xmath32 & @xmath33 & @xmath34 + fwhm ( mhz ) & @xmath35 & @xmath36 & @xmath37 & @xmath38 & @xmath39 + rms / mean ( % ) & @xmath40 & @xmath41 & @xmath42 & @xmath43 & @xmath44 ll @xmath8 & @xmath45 @xmath10 + @xmath46 & @xmath47 kev + @xmath48 & @xmath49 km + @xmath14 & @xmath50 + @xmath51 & @xmath52 + @xmath18 & @xmath53 + @xmath23 & @xmath54 + disk fraction & 27% + @xmath28 & 65.6/72 for the three chandra observations , the location of on the charge coupled device ( ccd ) was more than @xmath55 off the optical axis where a point source spreads on multiple pixels and photon pileup for sources at the flux level of is unimportant . event lists were created using ciao 4.1.2 with caldb 4.1.3 . source energy spectra in the 1 - 8 kev band were extracted from the source region and grouped by a factor of 4 in 1 - 4 kev , 8 in 4 - 6 kev , and 16 in 6 - 8 kev based on the original channels , with background subtracted from a nearby source - free region at the same location on the sky for the three observations . a power - law model subject to interstellar absorption provides adequate fits to the chandra spectrum . the absorption column density is almost constant at about @xmath56 @xmath10 . the source spectrum is hard , with a power - law photon index of about 1.3 - 1.4 . the luminosity corrected for absorption varied by a factor 1.5 among the three observations . all spectral parameters with 90% errors are listed in table [ tab ] . we tried to add an additional multicolor disk component to the power - law spectrum . pronounced improvement on the fits is only obtained for observation 8190 ; the disk component is favored at a significance of 4.1 @xmath29 . the best - fit model parameters are listed in table [ tab:8190 ] with 90% errors quoted . power spectra were calculated from lightcurves created using events from the same region for spectral analysis with a time step equal to the ccd frame time including the readout time , which is 0.44104 s for the three observations . each lightcurve was divided into @xmath57 segments with @xmath58 25 , 18 , and 8 for observations 6097 , 8190 , and 10027 , respectively . a fourier transform was performed for each segment and the individual power spectra were averaged and then geometrically binned by a factor of 1.2 , 1.1 , and 1.3 , respectively for the three observations . the binning factor is chosen for a reasonable frequency resolution around the qpos and a sufficiently large binning factor needed for the fitting . we checked different energy bands and found that the most significant timing noise appears in the 1 - 8 kev range . for the two xmm observations , the lightcurves were produced using merged pn and mos events in their common good time intervals ( gtis ) from a half circular region ( region b in fig . 2 of * ? ? ? * where the contamination from the other ulx is minimized ) with a time step 10 times the pn frame time of about 0.73 s. fourier transforms were performed in each gti with 4096 points , and the power spectrum was obtained by averaging all individual ones . the number of segments , @xmath57 , is 8 and 9 and the geometric binning factor is 1.1 and 1.2 , respectively for 0112290201 and 0560590101 . for observation 0560590101 , mos1 data were not included because they contain too many timing gaps . energy spectra of are not available from xmm data due to considerable photon confusion with nearby sources . spectral information for xmm 0560590101 can be obtained from the simultaneous chandra observation 10027 . for xmm 0112290201 , which does not have a joint chandra observation , the count rate of has been resolved from surface brightness fitting @xcite . there are four reliable spectral measurements of with chandra ( the three ones reported here and observation 5644 reported in * ? ? ? taking into account the spectral variations of with an absorption column density in the range of @xmath59 @xmath10 and a power - law index in 1.3 - 1.5 , the 1 - 8 kev luminosity of in xmm 0112290201 can be estimated using pimms as in the range of @xmath60 . prominent timing noise above the poisson level can be seen around a few mhz in the power spectra from these observations ( figure [ fig : pow ] ) , which we interpret as qpos . we fit the power spectra in the whole frequency range with a a gaussian plus constant model , which is overplotted on each spectrum with a dotted line . a lorentzian also fits these peaks , but not as well as the gaussian does . no obvious broadband red noise such as a ( broken ) power - law continuum is found near the qpo feature . we rule out the possibility that they are caused by instrumental oscillations like the dithering as follows . for the chandra observations , photons from nearby regions do not show similar oscillations and simulations with marx using the same aspect solution and bad pixel files confirms that no instrumentation signal is present at these frequencies . for the simultaneous chandra and xmm observations , the same qpo feature is detected by both . the xmm power spectra are substantially contaminated by photons from the nearby variable ulx , and the calculated qpo parameters may be affected . the photon fraction of in the extraction region is estimated and taken into account when calculating the rms amplitude of qpos from the xmm data . all qpo parameters are listed in table [ tab ] , and the qpo significance is calculated based on the @xmath61 variation on the normalization of the qpo component . we checked the power spectra at even lower frequencies for the three chandra observations , and found possible red noise only in 6097 appearing at frequencies well below the qpo frequency . for observation 8190 , the power spectrum between 0.01 - 1 mhz is dominated by the white noise . for observation 10027 , which is shorter than the other two , the powers between 0.1 - 1 mhz are also white noise dominated . there are several other chandra advanced ccd imaging spectrometer ( acis ) observations in which was active ( see * ? ? ? * ) but the mhz qpos were not detected . observation 5644 has an exposure of 70 ks and the source luminosity was @xmath62 in 1 - 8 kev . the number of photons from detected in 5644 is as many as in 6097 , and twice that in 10027 ; the nondetection of qpos is not due to low statistics . observation 2933 and 6361 both have an exposure of 18 ks and the source luminosity was @xmath63 and @xmath64 , respectively . another four acis observations ( 378 , 379 , 380 - 1 and 380 - 2 ) are shorter than 10 ks , and the source was slightly dimmer than in observation 2933 . the mhz qpos were not detected in any observation when the source was dim , below @xmath65 . low frequency qpos in stellar mass black hole x - ray binaries have a diverse population distributed in a wide frequency range . therefore , inference of the compact object mass via the qpo frequency should be done between qpos of the same type . the qpo coherence parameter @xmath66 , defined as the ratio of the central frequency to the full width at half maximum ( fwhm ) , is around 1.4 - 1.5 derived from the three chandra observations and nearly 4 from the xmm observation 0112290201 . the rms amplitude of the qpos in the three chandra observations is high , from 10% to 17% , but is relatively low in the xmm observation 0112290201 of 5% . we note that the chandra results should be more reliable due to less contamination . as mentioned above , possible red noise is only found in chandra 6097 , and the power spectrum is totally dominated by the white noise in the other two observations at frequencies down to 0.1 or even 0.01 mhz . therefore , the qpos in can be summarized as being broad and strong , and occurring with absent or weak red noise continuum . the qpos in are similar to four qpos detected in xte j1550@xmath1564 on 1999 mar 18 and 21 and apr 02 and 03 , respectively , which also have a low @xmath66 of about 1.0 , an rms amplitude of about 10% , and accompanied with a weak red noise @xcite . another similarity is that three out of the four qpos in xte j1550@xmath1564 do not exhibit a lorentz profile . these four qpos were classified as ` possible ' type a qpos by @xcite , or type a - i ( a subclass of type a ) by @xcite to distinguish them from weak type a qpos . @xcite argued that type a - i qpos should be classified as type b based on their frequency and amplitude . however , type b qpos are usually narrow , with a @xmath66 value of @xmath67 concluded from a few bhbs @xcite . therefore , the type a - i could be a special subclass amid type a and type b. in spite of the controversy over the classification of type a - i , we identify the qpos found in as of type a - i based on their similar profile and the weak / absent red noise , or conservatively , of type a / b according to the discussion above . as noted above , the mhz qpos were detected only when the source was luminous , above @xmath68@xmath0 . type a / b qpos show identical behavior that they appear only when the source flux is high enough @xcite . this further supports the identification of the mhz qpos as of type a / b . by any means , the qpos in are not analogous to type c , which are narrow , strong , and accompanied by a strong red noise continuum ; otherwise , the following discussion would be different as type c qpos appear at a frequency range significantly wider than others . the four type a - i qpos in xte j1550@xmath1564 have a central frequency varying between 7.1 to 10.3 hz ; all type a qpos in xte j1550@xmath1564 are found in a frequency range of 4.88 - 10.3 hz , which covers the frequency range of type b qpos ( 4.94 - 6.1 hz ) in the same source @xcite . the qpos in have a central frequency between 2.77 - 3.98 mhz . comparing with the type a - i qpos in xte j1550@xmath1564 , the upper and lower bounds of the frequency range have exactly the same ratio but scaled by a factor of about 2600 . based on the assumption of type a - i qpos , the compact object mass of can be inferred as 25,000 - 30,00 @xmath2 after adopting a mass of 9.68 - 11.58 @xmath2 for xte j1550@xmath1564 @xcite . a more conservative estimate taking into account all type a / b qpos leads to a black hole mass of 12,000 - 43,000 @xmath2 in , inferred from the possibly smallest and largest ratios between qpo frequencies from the two sources . type a / b qpos are detected in several other bhbs besides xte j1550@xmath1564 , including gx 339@xmath14 , xte j1859 + 226 , and gs 1124@xmath1684 ( see * ? ? ? * and references therein ) . in these sources , the type a / b qpos show up in a similar frequency range , e.g. 4.5 - 7.49 hz in gx 339@xmath14 @xcite and 4.45 - 8.42 hz in xte j1859 + 226 @xcite . type b qpos are also detected in grs 1915 + 105 varying in the frequency range of 2.44 - 6.84 hz @xcite . they are broad but shown above a strong red noise component , partially similar to the qpos in . these sources also contains a similar massive black hole as in xte j1550@xmath1564 , but their masses are not as well constrained as in the latter @xcite . therefore , inferring the mass of via type a / b qpos from these sources will lead to consistent but looser constraints . we note that all above estimates of the mass depend on a hypothesis that the frequency of type a / b qpos is linearly scaled with the black hole mass , which , however , has never been tested . more work regarding the mass dependence of type a / b qpos is needed . in gx 339@xmath14 , all type a / b qpos were detected when the source was in the soft intermediate state @xcite , also known as the steep power - law state defined by @xcite . in xte j1550@xmath1564 , the four type a - i qpos were found in the steep power - law state as well @xcite . however , the x - ray spectrum of is hard , seemingly inconsistent with the steep power - law state . disk emission is detected in observation 8190 , with a moderate fraction of 27% in 1 - 8 kev . this disk component will contribute 40% to the total emission below 1 mev if the power - law component is not cut off . in the hard state , the disk contributes less than 25% in 2 - 20 kev for stellar mass black holes . for more massive black holes , though there is no defined energy range , a disk fraction of 27%-40% seems too high for being classified as in the hard state . a varying disk component and an energetic power - law component suggest that the source may be in a state similar to the steep power - law state in stellar mass black holes . interestingly , the best - fit disk inner radius is consistent with the gravitational radius of a black hole of 11,000 - 44,000 @xmath2 , which is in agreement with the mass estimated from the qpo frequency . lies about @xmath4 from the kinematic center of m82 @xcite on the sky plane . following the discussion in @xcite and references therein , an upper limit of the compact object mass can be placed of about @xmath69 @xmath2 such that the black hole will not fall onto the nucleus of the galaxy due to dynamical friction if the black hole was born at the same time of the galaxy . this upper limit is consistent with our mass estimate via qpos . is coincident with an extended radio source known as 42.21 + 59.2 , which is believed to be an h ii region @xcite associated with a star cluster seen from infrared @xcite . therefore , it could be ruled out that the source is a background agn . if the disk component is true , its high bolometric luminosity of @xmath70 and low temperature make it a good candidate for the ionization source of the h ii region . the outburst luminosity of the source is about @xmath0 . in the total 20 observations of the source ( 16 reported in * plus chandra 8190 , 10027 , 10025 , and 10026 ) , was found active in 12 of them , indicative of a duty cycle @xmath71 . assuming an accretion efficiency @xmath72 , the mean mass accretion rate of the source is found to be @xmath73 g s@xmath6 . if the source is accreting from the interstellar medium directly , the mass accretion rate in g s@xmath6 can be written as @xmath74 where @xmath57 is the black hole mass , @xmath75 is density and @xmath76 is the speed of sound of the medium @xcite . adopting canonical values of the density and sound speed shown in the equation , a black hole mass of @xmath77 @xmath2 is required to power the x - ray luminosity without a companion star . this means that if the mass estimated from the qpos are reliable , the source does not have to be in a close binary system , but is massive enough to accrete from the interstellar medium directly . @xcite suggests that the bondi rate may overestimate the true accretion rate for agns by 1 - 2 orders of magnitude . in this case , a corresponding higher black hole mass would be required . conversely , if the ulx indeed lies within a star cluster , the gas density could be substantially higher , which would decrease the required black hole mass . anyway , one should be cautious of the high uncertainty on such an estimate . we thank the anonymous referee for helpful comments and the mission planning teams of chandra and xmm - newton for making the observations possible . hf acknowledges funding support from the national natural science foundation of china under grant no . 10903004 and 10978001 , the 973 program of china under grant 2009cb824800 , and the foundation for the author of national excellent doctoral dissertation of china under grant 200935 . pk acknowledges partial support from chandra grant go9 - 0034x and nasa grant nnx08aj26 g .	we report the discovery of x - ray quasi - periodic oscillations ( qpos ) at frequencies of 3 - 4 mhz from a transient ultraluminous x - ray source ( ulx ) in m82 . the qpos are strong and broad and appear with weak or absent red noise , and are detected only in chandra observations when the source is brighter than @xmath0 . the qpo behavior is similar to the type a - i qpos found in xte j1550@xmath1564 , which is a subclass of low frequency qpos with properties in between type a and b. therefore , we identify the qpos in as of type a or b , and rule out the possibility of type c. with this identification , the mass of the black hole in can be inferred as in the range of 12,000 - 43,000 solar masses by scaling the qpo frequency to that of the type a / b qpos in stellar mass black holes . cool disk emission is detected in one chandra observation , and the disk inner radius suggests a similar black hole mass range . black holes of such a high mass are able to produce an energy output in a manner similar to by accreting from the interstellar medium directly .
in recent years much attention has been paid to the emerging field of spintronics in semiconductor physics . much of the interest in this field is stimulated by the spin - field - effect transistor proposed by datta and das @xcite . in this proposal the rashba spin - orbit interaction is used to manipulate the charge transfer through a transistor like device by controlling the electron spin . the rashba interaction , which is the conventional spin - orbit interaction with a constant electric field , affects the electron spin like a momentum dependent magnetic field . in practice the realization of the proposal by datta and das has proven to be a challenging task . in particular the injection of spins into non - magnetic semiconductors has turned out to be demanding , although successful injection has been reported by several authors @xcite . despite this fact the importance of the rashba interaction for the concept of the spin - field effect transistor has motivated further studies on its impact on charge and spin transport properties . particular attention has been paid , e.g. , to investigations of spin - accumulations@xcite and to the spin - galvanic effect@xcite , to an intrinsic spin - hall effect@xcite , and to the weak localization - antilocalization transition in the conductivity @xcite . since in the rashba model the electron wave - vector is coupled to the spin the spin is quickly randomized by scattering events . therefore , the ballistic transport regime is the most important regime for the spin - field effect transistor proposed by datta and das . accordingly , most of the investigations focus on the situation in which the disorder energy is small compared to the rashba level splitting . recently , however , a generalized spin - field effect transistor has been proposed@xcite which is also expected to work in the diffusive regime , even if the rashba level splitting is small compared to the characteristic disorder energy . this raises the question which impact of the rashba interaction on the charge and spin transport properties can be expected in this limit . the aim of the paper is to provide a theory for the spin - relaxation in a a two - dimensional electron gas in the diffusive regime , in the limit that the characteristic disorder energy is large compared to the rashba level splitting on a time scale , in which inelastic processes can be ignored . to this end we generalize the non - nonlinear @xmath0-model to include also the impact of the rashba interaction and an in - plane electric field . using the generalized @xmath0-model we derive generalized bloch - equation and use these equations for the investigation of an initial magnetization . we consider charge carriers in a two - dimensional plane . the motion of the charge carriers is described by the hamilton operator @xmath1 here @xmath2 is the momentum operator , @xmath3 is the effective mass , @xmath4 is a constant field in the two - dimensional plane , @xmath5 is a constant vector perpendicular to the plane , and @xmath6 is a random potential . the components of the vector @xmath7 are the pauli matrices . @xmath8 and @xmath9 are the unit vectors in the two - dimensional plane and @xmath10 is a unit vector transverse to the plane . accordingly , @xmath11 and @xmath12 . the random potential is characterized by a gaussian distribution function with zero mean and second moment @xmath13 here @xmath14 denotes the configuration average , @xmath15 is the density of states per spin at the fermi - level for @xmath16 , and @xmath17 is the single - particle scattering time . in our investigation we focus on configuration averages of retarded and advanced green functions . these functions are solutions to the differential equation @xmath18 in this equation @xmath19 is a @xmath20 matrix in spin - space , @xmath21 is the @xmath20 unit matrix and @xmath22 and @xmath23 are fixed energies . as a boundary condition we require that the green functions vanish at infinity . in this case the configuration averaged green functions are invariant against translations by a fixed vector @xmath24 if the energy @xmath22 is also shifted by @xmath25 . accordingly , they satisfy the relationship @xmath26 to calculate the configuration averages of the green functions we use the field theoretical formulation of the refs . [ ] . in this formulation the replica trick is used to derive an effective action for the calculation of the configuration averaged green functions . following the standard derivation we find that for the problem at hand the partition function takes the form @xmath27 where @xmath28 here @xmath29 is a hermitian @xmath30 matrix field ( @xmath31-number of replica copies+ 2 components particle - hole space+ 2 components spin indices ) with components @xmath32 ( @xmath33- replica indices , @xmath34- particle - hole indices , @xmath35-spin indices ) . it satisfies the relationship @xmath36 , where @xmath37 is the transposed matrix and @xmath38 the tr in eq.([b6 ] ) contains both the summation over the discrete indices and the integration over space . furthermore , @xmath39\delta({\bm x}-{\bm x'}),\end{aligned}\ ] ] where @xmath40 , @xmath41 for @xmath42 and @xmath43 for @xmath44 . the index @xmath45 can take on values between @xmath46 and @xmath31 . it is understood that the limit @xmath47 is taken at the end of the calculation . to calculate the configuration - averaged green function in saddle - point approximation we now look on the saddle - point solution of the effective action @xmath48 , defined by @xmath49 the saddle - point values of @xmath29 and @xmath50 are diagonal with respect to the replica indices , as are the exact expection values . accordingly , @xmath51 and @xmath52 . from the eqs . ( [ b4 ] ) and ( [ b6 ] ) we find @xmath53 here @xmath54 . to find an equation for the green function we use the gradient expansion @xcite . to this end we first introduce wigner coordinates , according to @xmath55 and @xmath56 , and perform a fourier transformation with respect to the relative coordinates . the green function in the new coordinates @xmath57 is related to the old green function by the relationship @xmath58 in calculating @xmath57 we restrict the consideration to the lowest order in the gradient expansion . in this approximation the equation for the green function takes the simple form @xmath59\delta_{ij}+i\frac{\hbar}{2\tau } { \tilde q}_{\alpha}^{ij}(\mu_r ) ) { \tilde g}_{\alpha}^{jk}({\bm r } , { \bm k}\|e,\omega ) = { \bm 1}\delta_{ik}\end{aligned}\ ] ] here all objects are considered as @xmath60 matrices in spin space , @xmath61 and @xmath62 . in order to solve the eqs.([sp2])-([sp4 ] ) self - consistently we restrict the consideration to the classical accessible region @xmath63 . furthermore , we consider only the dirty limit , in which both the disorder energy @xmath64 and @xmath65 are large compared to the rashba level splitting @xmath66 and assume that @xmath65 is large compared to @xmath64 . in this case we find that the self - consistent solution to the equations ( [ sp2 ] ) and ( [ sp4 ] ) takes the form @xmath67 @xmath68 @xmath69 @xmath70 where @xmath71 and @xmath72 . in writing down the saddle - point field ( [ sp5 ] ) we have ignored its imaginary part which only leads to an @xmath45-independent shift of @xmath65 . we would like to note that the simple structure of the saddle - point field is a consequence of the fact that the density of states for @xmath16 is constant . in an energy dependent density of states the saddle - point would pick up an energy and field dependent contribution from the density of states , which would modify eq.([sp5 ] ) . the fluctuations around the saddle - point solution are characterized by massive and soft modes . the latter describe the diffusion and spin relaxation properties we are interested in . in the field - theoretical formulation the existence of these modes is a consequence of a continuous symmetry of the effective action ( [ b6 ] ) which exists in the limit @xmath73 . in this limit the effective action is invariant against similarity transformations with hermitian matrices @xmath74 , which satisfy the relationship @xmath75 such matrices are generated by hermitian generators @xmath76 ( @xmath77 ) , which have the structure @xmath78\nonumber\\ & & + { s_{\alpha\alpha'}}_0 ^ 3\sigma_3\otimes\sigma_0+i { a_{\alpha\alpha'}}_i^3\sigma_3\otimes\sigma_i\end{aligned}\ ] ] here @xmath79 ( @xmath80 ) are real matrices symmetric ( antisymmetric ) in @xmath45 and @xmath81 , @xmath82 , and @xmath83 corresponds to @xmath84 , @xmath85 and @xmath86 , respectively . the components of the matrices @xmath87 , which are associated with the upper index of the matrices @xmath79 and @xmath80 ( that is with @xmath88 ) , correspond to the particle - hole indices . the lower indices characterize the spin . in order to derive an effective action for the soft modes we follow the standard derivation for the non - linear sigma - model @xcite . the complications originating from the presence of the electric field we treat as in ref . [ ] . doing so , we find that the effective action for the diffusion and spin relaxation modes takes the form @xmath89 here @xmath90 is the soft part of the matrix @xmath91 generated from the saddle - point field @xmath92 by the local similarity transformation @xmath93 , @xmath94 and @xmath95.\ ] ] the bracket @xmath96 $ ] symbolizes the commutator and the step function @xmath97 restricts the range of integration to the classical accessible region . the matrix @xmath98 in eq.([sm5 ] ) is a matrix which only acts on the spin - indices but not on the replica and the particle - hole indices . in order to investigate the non - linear @xmath0-model further we write the matrix @xmath99 in the form @xmath100 where @xmath101 @xmath42 , @xmath102 and @xmath103 , where + symbolizes hermitian conjugation , and expand the action ( [ sm3 ] ) with respect to powers of @xmath104 up to fourth order in @xmath104 . doing so , we use the fact that the matrices @xmath105 are antisymmetric with respect to @xmath45 and @xmath81 for @xmath106 and @xmath107 and for @xmath108 and @xmath109 and symmetric with respect to @xmath45 and @xmath81 for @xmath108 and @xmath107 and for @xmath106 and @xmath109 . the quadratic terms in the expansion yield the gaussian propagator , the physics of which agrees with the ladder - approximation . the quartic terms in the expansion yield the weak - localization corrections to the transport coefficients . in the gaussian approximation the effective action takes the form @xmath110 where @xmath111 for @xmath106 and @xmath109 and for @xmath108 and @xmath107 and @xmath112 else . the operator @xmath113 , the non - vanishing components of which are @xmath114 and @xmath115 is related to the generalized diffusion propagator @xmath116 by the relationship @xmath117 here the bracket @xmath118 symbolizes the anti - commutator , @xmath119 and the indices @xmath120 and @xmath121 can take on the values @xmath122 , which correspond to @xmath84 , @xmath85 , and @xmath86 , respectively . the transport coefficients are given by @xmath123 @xmath124 and @xmath125 . according to the eqs.([ga2 ] ) , ( [ ga3 ] ) and ( [ ga4 ] ) the equation for the particle transport decouples from the equations for the spin transport . the particle transport is described by the propagator @xmath126 . its equation of motion contains besides ordinary particle diffusion also the particle packet drift and the heating of the charge carriers . its physics is detailed in ref . the spatial components of the diffusion propagator , @xmath127 for @xmath128 , describe spin diffusion and relaxation . in addition to the diffusion , drift and heating processes present in the particle number diffusion equation , these equations also contain the spin relaxation and an electric - field dependent rotation of the spin - direction , which is induced by the rashba - interaction . we would like to note that the same equations can also be obtained from the ladder approximation . in order to calculate the quantum corrections to the transport coefficients we also take into account the quartic terms in the expansion of @xmath129 with respect to @xmath104 and investigate the corresponding functional integral in one - loop approximation . doing so , we restrict the consideration to the linear response regime . in this case we can ignore the impact of the electric field on the transport coefficients and thus set the external electric field @xmath4 equal to zero in all loops . for charge transport the range of validity of this approximation has been discussed in ref . [ ] . there it has been shown that the impact of the field is negligible , as long as the spread of the particle packet is small compared to the distance of its center from the turning point . the time needed by a particle packet , which was a delta - pulse centered at @xmath130 at time @xmath131 and in which all particles have the same energy @xmath22 , to reach a width of the order of @xmath132 is for a typical experiment larger than @xmath133 s , as shown in ref . given that the typical lifetime of an initial magnetization is in the nanosecond range it seems to be reasonable to restrict the consideration to this limit . to one - loop order the particle number diffusion coefficient @xmath134 in eq.([ga2 ] ) is replaced by @xmath135 where @xmath136 @xmath137 and @xmath138 here @xmath139 is the ultraviolet cutoff . these corrections have the same structure as the conventional anti - weak localization corrections to the conductivity @xcite . in a current relaxation experiment the frequency @xmath140 is effectively replaced by @xmath141 , where @xmath142 is the inelastic phase breaking time which increases with decreasing temperature . accordingly , the particle number diffusion coefficient has a minimum as a function of the temperature at @xmath143 , where @xmath144 . therefore , the diffusion coefficient increases with decreasing temperature if @xmath145 . the opposite turns out to be true for the spin - transport coefficients . to one - loop order the spin diffusion coefficients @xmath146 and @xmath147 are replaced by @xmath148 @xmath149 by @xmath150 the spin relaxation frequencies @xmath151 by @xmath152 and @xmath153 and the rotation frequency @xmath154 by @xmath155 in these equations @xmath140 has also to be replaced by @xmath141 if @xmath140 is smaller than @xmath141 . in contrast to the particle - number diffusion coefficient , however , the spin - transport coefficients have no minimum as a function of the temperature . in the range of applicability of the eqs.([wl5])-([wl9 ] ) the magnitude of the spin transport coefficients keeps decreasing with decreasing temperature even if @xmath145 . in contrast to the corrections to the particle- number diffusion coefficient , which are anti - localizing at low temperatures , the corrections to the spin transport coefficients always show a tendency to localization . at this point we would also like to mention that the quantum corrections render the spin - transport coefficients weakly anisotropic . this anisotropy results from the fact that @xmath156 . however , since the difference between these quantities is of the order of @xmath157 this difference is small compared to the magnitude of the quantum corrections themselves , which are larger than the former by a factor of the order of @xmath158 . therefore , we ignore the anisotropy in the following and replace @xmath159 by @xmath160 everywhere . in this case @xmath161 and @xmath162 the evolution of an initial spin packet is described by the bloch equations . phenomenology tells us that the bloch equations have the structure @xmath163 here @xmath164 is the laplace transform of the spin density with respect to time , @xmath140 is the corresponding laplace frequency , @xmath165 is the initial condition , @xmath166 is a second rank tensor , @xmath167 is a vector , and @xmath168 is the spin - current tensor . for the components we use the notation @xmath169 and @xmath170 . the quantities @xmath164 , @xmath165 and @xmath168 depend also on @xmath171 and on the total energy @xmath22 , but this dependence has been suppressed in eq.([dmm1 ] ) to simplify the notation . due to the generalized translation invariance ( [ b4 ] ) the transport coefficients @xmath172 and @xmath173 do not depend on the quantities @xmath171 and @xmath22 separately , but only on the kinetic energy @xmath174 . in eq.([dmm1 ] ) this dependence has also been suppressed to simplify the notation . from the physical point of view the tensor @xmath175 describes the decay of the initial magnetization and @xmath167 describes the rotation of the initial magnetization . the disorder renders both quantities dispersive . therefore , both @xmath176 and @xmath167 depend on @xmath140 . this dependence should be observable in optical experiments where it should reflect itself in deviations from the purely exponential decay . the equations ( [ ga3 ] ) and ( [ ga4 ] ) can be cast into the form ( [ dmm1 ] ) . doing so , it has to be taken into account that the bare transport coefficients are renormalized by the one - loop corrections of the preceding section.from eq.([ga3 ] ) we find @xmath177 and @xmath178 the spin current tensor takes the form @xmath179 this tensor has been obtained by collecting all those terms in eq.([ga3 ] ) , which can be cast into a divergence . accordingly , the tensor ( [ dmm4 ] ) might deviate from the true spin - current tensor by terms which are annihilated by the divergence . however , since every observable quantity can be calculated directly from eq.([dmm1 ] ) such terms are not important for our further considerations . the eqs.([wl1])-([wl9 ] ) ignore the impact of inelastic scattering events on the transport coefficients . to take them into account we introduce the phenomenological phase relaxation time @xmath142 and replace @xmath180 . in this case the transport coefficients are given by the eqs.([dmm2 ] ) and ( [ dmm3 ] ) with @xmath140 replaced by @xmath141 for @xmath181 . accordingly , the relaxation becomes markovian for @xmath182 . below we restrict the consideration to this situation . in order to investigate some of the consequences of eq.([dmm1 ] ) we first focus on the decay of a spatially homogeneous magnetization . to this end we assume that an initial magnetization @xmath165 has been created at time @xmath131 and that only charge carriers in the vicinity of the fermi - energy @xmath183 contribute to the magnetization . in this case the initial distribution function is given by @xmath184 in the absence of the field . in the presence of the field the energy of a charge carrier placed at @xmath171 is changed by @xmath185 . therefore , the initial condition takes the form @xmath186 in the presence of the field . during the evolution the spin distribution remains spatially homogeneous . accordingly , @xmath187 . the direction and the magnitude of the magnetization and the energy of the particles are changed . since , however , the magnetization decays on a time scale small compared to the time needed by a particle to change its energy appreciably the energy distribution function remains close to the initial delta - function during the time in which the magnetization is measurable . to take advantage of this fact we make the ansatz @xmath188 and derive a closed equation for @xmath164 by integrating eq . ( [ dmm1 ] ) with respect to @xmath174 . doing so , we obtain the simple algebraic system of equations @xmath189_{x / y}=s^0_{x / y},\ ] ] @xmath190_{z}=s^0_{z}.\ ] ] this system has the same structure as that for the decay of an initial magnetization in an ordered system ( hopping transport of small polarons on an ordered lattice ) @xcite . only the structure of the transport coefficients is different . for @xmath191 ( solid line ) , 2 ( dashed line ) and 10 ( dashed dotted line ) . the initial conditions are @xmath192 and @xmath193.,width=359 ] for @xmath192 and @xmath193 . the parameters are the same as in fig.1.,width=359 ] to calculate the magnetization we solve this system and perform an inverse laplace transformation . doing so , we obtain @xmath194 @xmath195\nonumber\\ & & + s_z^0\omega_f\frac{\sinh(\omega_f t ) } { \omega_f } ] % \end{aligned}\ ] ] and @xmath196\nonumber\\ & & - s_x^0\omega_f\frac{\sinh(\omega_f t ) } { \omega_f } ] . % \end{aligned}\ ] ] here @xmath197 , @xmath198 and @xmath199 the eqs.([dm8])-([dm10 ] ) apply for @xmath182 . they do not contain the magneto - electric effect of ref . [ ] , since this effect is proportional to the ratio between the rashba level splitting and the fermi energy , which is considered as small in our treatment . according to the eqs.([dm8])-([dm11 ] ) the character of the decay of the initial magnetization depends on the magnitude of the electric field . if @xmath200 , where @xmath201 is determined from the requirement @xmath202 , the frequency @xmath203 is real and reflects itself in an exponential decay of an initial magnetization ( see the figs.1 and 2 ) . furthermore , the @xmath84-component of the magnetization is coupled with the @xmath86-component of the magnetization . if @xmath204 a non vanishing magnetization in @xmath84-direction is created even if @xmath205 . if @xmath206 the frequency @xmath203 becomes purely imaginary . in this case the hyperbolic functions turn to trigonometric functions , which lead to a rotation of the decaying initial magnetization ( see fig.1 and 2 ) . therefore , the vector @xmath207 , with components @xmath208\nonumber\\ & & + s_z^0\omega_f\frac{\sin(\|\omega_f\| t ) } { \|\omega_f\| } ] % \end{aligned}\ ] ] and @xmath209\nonumber\\ & & - s_x^0\omega_f\frac{\sin(\|\omega_f\| t ) } { \|\omega_f\| } ] , % \end{aligned}\ ] ] moves on an ellipse with decaying diameter . the ellipse satisfies the equation @xmath210 the quadratic form can be diagonalized . doing so , it can be shown that the elliptic axis are tilted by @xmath211 . the rotation of the magnetization discussed in the preceding subsection reflects also in the current . in the presence of the rashba interaction the current operator has the form@xcite @xmath212 here @xmath213 is the charge and @xmath214 is the spin operator . the heisenberg equation for @xmath215 is independent of @xmath216 . accordingly , the configuration average of the expection value of the first term on the right hand side of eq.([ah1 ] ) yields the conventional drude current and its weak localization corrections . the configuration average of the second term yields two additional contributions , a correction term to the longitudinal current @xmath217 and the hall - like term @xmath218 perpendicular to both @xmath4 and @xmath219 . if we chose the same initial condition for the magnetization as in the preceding section @xmath220 is oscillating , if @xmath221 . according to eq.([ah3 ] ) these oscillations manifest themselves in an alternating hall - like current.the alternating hall current is produced by a constant electric field . however , since the magnetization decays when time elapses the current oscillations are not stable but decay too . this raises the interesting question whether these oscillations can be stabilized by some means . in the paper we have used field theoretical methods to derive the generalized bloch equations for the magnetization in a disordered two - dimensional electron gas . our equations apply for time scales which are large compared to the elastic scattering time , large compared to the phase - breaking time but small compared to energy relaxation time . accordingly , they take into account the particle- and spin diffusion , the drift induced by an external electric field in the plane , the heating of the charge carriers , the decay of the magnetization due to scattering events , and an electric field dependent rotation of the magnetization direction . the calculation of the transport coefficients shows that the impact of the quantum - corrections on the spin transport coefficients is different from that on the particle transport coefficients . the corrections to the latter are antilocalizing . this fact manifests itself in a minimum of the diffusion coefficient as a function of the temperature . the structure of the corrections to the particle transport coefficients agrees with the conventional antilocalization corrections to the conductivity due to spin - orbit coupling@xcite . in contrast to the particle transport coefficients the spin transport coefficients show a tendency to localization . both the spin diffusion coefficient and the spin decay rate decrease with decreasing temperature , even at very low temperatures . this sets the situation for the spin - transport coefficients apart from that for the particle transport coefficients , which increase with decreasing temperature at sufficiently low temperature due to antilocalization . we have applied these equations to an investigation of the decay of a spatially homogeneous initial magnetization . in the absence of the in - plane electric field such a magnetization decays exponentially at large times@xcite . the decay is anisotropic . its transverse component decays twice as fast as its longitudinal components.the temperature dependence of the decay is given by the temperature dependence of the weak localization corrections to the transport coefficients , which lead to a logarithmic reduction of the decay rate with decreasing temperature . accordingly , the decay becomes slower with decreasing temperature . already a weak electric field renders the decay qualitatively different . if the electric field is smaller than a critical field it tips the magnetization only slightly over . if , however , the field exceeds the critical field @xmath201 the magnetization starts rotating in the plane transverse to @xmath222 . the rotation frequency depends on the degree of disorder , on the magnitude of the rashba - parameter , on the magnitude of the electric field and on temperature . an increase of the temperature results also in an increase of the rotation frequency @xmath203 and in an increase of the critical field @xmath201 . we would like to note that a similar rotation has also been obtained in ref . [ ] for small polarons on an ordered lattice . our calculation shows that this rotation can also be observed in a disordered electron gas , in which the transport mechanism is very different from that for a small polaron on an ordered lattice . this work was done during a stay at the the university of oregon . the author is grateful for the hospitality of the material science institute , in particular to d. belitz , without whose support and encouragement this work would not have been accomplished . i would also like to thank v. v. bryksin for helpful discussions . this work was supported by the dfg under grant . bl456/3 - 1 . s. datta and b. das , appl . phys . lett . * 56 * , 665 ( 1990 ) . r. fiederling , m. keim , g. reuscher , w. ossau , g. schmidt , a. waag , and l. w. molenkamp , nature ( london ) * 402 * , 787 ( 1999 ) . y. ohno , d. k. young , b. beschoten , f. matsukura , h. ohno and d. d. awschalom , nature ( london ) * 402 * , 790 ( 1999 ) . h. j. zhu , m. ramsteiner , h. kostial , m. wassermeier , h. p.schonherr and k. h. ploog , phys * 87 * , 016601 ( 2001 ) . g. schmidt , g. richter , p. grabs , c. gould , d. ferrand and l. w.molenkamp , phys . lett . * 87 * , 227203 ( 2001 ) . p. r. hammar and m. johnson , phys . rev . lett . * 88 * , 066806 ( 2002 ) . jun - ichiro inoue , g. e. w. bauer and l. w. molenkamp , phys . rev . b * 67 * , 33104 ( 2003 ) . s. d. ganichev , e. l. ivchenko , v. v. belkov , s. a. tarasenko , m. sollinger , d. weiss , w. wegschneider and w. prettl , nature ( london ) * 417 * , 153 ( 2002 ) . j. sinova , d. culcer , q. niu , n. a. sinitsyn , t. jungwirth , and a. h. macdonald , cond - mat + 0307663 . i. l. aleiner and v.i . falko , phys . rev . lett . * 87 * , 256801 ( 2001 ) . p. w. brouwer , j. n. h.j . cremers , and b. i. halperin , phys.rev.b65 , 081302(r ) ( 2002 ) . d. m. zumbhl , j. b. miller , c. m. marcus , k. campman , and a. c. gossard , phys . lett . * 89 * , 276803 ( 2002 ) . takaaki koga , junsaku nitta , tatsushi akazaki , and hideaki takayanagi , phys . rev . lett . * 89 * , 46801 ( 2002 ) . j. b. miller , d. m. zumbhl , c. m. marcus , y. b. lyanda - geller , d. goldhaber - gordon , k. campman , and a. c. gossard , phys . rev.lett.90 , 76807 ( 2003 ) . j. schliemann , j. c. egues and d. loss , phys . * 90 * , 146801 ( 2003 ) . a. j. mckane and m. stone , ann . ( ny ) * 131 * , 36 ( 1981 ) . a. m. m. pruisken and l. schfer , nucl . b * 200 * [ * fs4 * ] , 20 ( 1982 ) . d. belitz and t. r. kirkpatrick , rev . phys . * 66 * , 261 ( 1994 ) . l. kadanoff and g. baym , _ quantum statistical mechanics _ , addison - wesley pub . o. bleibaum and d. belitz , submitted to phys . b. ( cond - mat/0309470 ) . s. hikami , a. i. larkin and y. nagaoka , prog . . phys . * 63 * , 707 ( 1980 ) . t. damker , h. bttger and v. v. bryksin , phys . stat . sol.(b ) in print . v. m. edelstein , solid state . 73 * , 233 ( 1990 ) . l. w. molenkamp , g. schmidt and g. e. w. bauer , phys . b * 64 * , 121202(r ) ( 2001 ) . u. zlicke and c. schroll , phys . rev . lett . * 88 * , 29701 ( 2002 ) . the magneto - electric effect of ref . [ ] is absent in our approach since we have ignored the rashba level splitting compared to the fermi energy .	the impact of an external electric field on the spin relaxation in a disordered , two - dimensional electron system is studied within the framework of a field - theoretical formulation . generalized bloch - equations for the diffusion and the decay of an initial magnetization are obtained . the equations are applied to the investigation of spin relaxation processes in an electric field .
the theory of medium - induced gluon radiation has received a lot of attention over the last twenty years , as it became clear that parton energy loss is most likely responsible for the quenching of large @xmath5 hadrons and jets in heavy - ion collisions , first observed at rhic @xcite and then at the lhc @xcite . multiple scatterings of a high - energy parton propagating in a medium applies more naturally to the former case . ] of length @xmath6 induce the emission of gluons , a process occurring on a typical time @xmath7 , where @xmath8 is the light - cone momentum of the radiated gluon , and @xmath9 . it is useful to recall the three regimes identified in @xcite depending on the value of the gluon formation time @xmath10 ( for a more detailed discussion in qed and in qcd , see @xcite ) : * * bethe heitler , @xmath11 * , where @xmath12 is the parton mean free path in the medium . in this regime , _ each _ scattering center acts as an independent source of radiation . the induced gluon spectrum is thus proportional to the ( bethe - heitler ) gluon spectrum induced by a _ single _ scattering . this regime is at work for very soft gluons , @xmath13 , where @xmath14 is the typical transverse momentum exchange in a single scattering . * * landau pomeranchuk migdal ( lpm ) , @xmath15 * , for which a _ group _ of @xmath16 scattering centers acts as a single radiator , leading to a relative suppression of the gluon radiation intensity with respect to that in the bethe - heitler regime . the above condition on @xmath10 translates into @xmath17 , where @xmath18 is the so - called medium transport coefficient . ( in a large medium , @xmath19 turns out to be on the order of @xmath20 , the transverse momentum exchanged over the length @xmath6 . ) * * long formation time , @xmath21. in this regime ( also known as the ` factorization regime ' ) , _ all _ scattering centers in the medium act coherently as a source of radiation . it occurs at large gluon momenta , @xmath22 . the region of large formation time leads to ` large ' mean parton energy loss , @xmath23 ( with @xmath24 the parton light - cone momentum ) , coming from the broad time interval for gluon radiation . however , because of the weak dependence of @xmath25 on the medium length , it is often considered that radiation of gluons with long formation times does not contribute to the _ medium - induced _ energy loss . as a consequence , gluon radiation in the lpm regime would dominate the induced spectrum , leading to an average parton energy loss @xmath26 , independent of @xmath24 . however , it was shown in ref . @xcite that the latter holds when the energetic parton is suddenly created ( or annihilated ) in the medium , but _ does not hold _ in the case of an ` asymptotic ' parton prepared long before the medium , undergoing small angle scattering through the medium , and hadronizing long after . in this case , the explicit calculation shows that the medium - induced spectrum is dominated by gluons with large formation time @xmath27 , being thus _ fully coherent _ over the medium . this leads to an induced parton energy loss proportional to the parton energy , @xmath23 . ( in particular , this demonstrates that the induced energy loss in a finite length medium is not bounded at high energy ) . in light - cone gauge , the _ medium - induced _ coherent radiation spectrum arises from the interference between gluon emission in the ` initial state ' ( _ i.e. _ , before the medium ) and in the ` final state ' ( after the medium ) @xcite . as discussed in ref . @xcite , coherent parton energy loss is expected to play an important role in the phenomenology of hadron production in proton - nucleus collisions . the effects of coherent energy loss on quarkonium nuclear suppression were studied in refs . @xcite . the kinematical setup used in @xcite where fully coherent radiation occurs is as follows : a parton of high energy ( large @xmath24 ) and with @xmath28 experiences a single hard scattering ( with transverse momentum exchange @xmath29 ) off a nuclear target and is tagged with @xmath30 at small angle @xmath31 . in addition to such @xmath32 _ hard _ and _ forward _ scattering , the parton undergoes multiple _ soft _ scattering in the finite - size target , and thus receives a transverse momentum kick @xmath33 . in this setup , ref . @xcite ( see also @xcite ) considered the @xmath34 hard process ( mediated by single gluon exchange in the @xmath35-channel ) , with the final @xmath36 pair being massive ( of mass @xmath37 ) and in a compact color octet state . the associated induced coherent spectrum was derived in a feynman diagram calculation and at first order in the opacity expansion , _ i.e. _ , with the transverse momentum broadening @xmath38 across the medium modelled by a _ single _ rescattering @xmath39 . the result at all orders in the opacity expansion was obtained by identifying @xmath40 ( where @xmath41 , with @xmath42 the gluon mean free path in the medium ) . the induced spectrum found in this semi - heuristic way reads @xcite [ spec - aps-0 ] x = n_c , when defined with respect to a zero - size target , or equivalently [ spec - aps ] . x \|_pa - pp = n_c , when defined in p a vs. p p collisions . we denote @xmath43 and @xmath44 . medium - induced coherent radiation was also studied using a semi - classical method in refs . @xcite , at first order @xcite and all orders @xcite in opacity and in a similar kinematical setup as that of ref . @xcite , however for the @xmath2 case of a massless quark experiencing a hard scattering mediated by a _ color singlet _ exchange in the @xmath35-channel . although no compact form of the coherent energy spectrum is given in refs . @xcite , the integral over @xmath45 of the @xmath45-differential spectrum found in @xcite can be worked out analytically and reads ( with @xmath46 , @xmath47 ) [ spec - armesto2 ] . x be checked to have the same parametric limits as the spectrum ( [ spec - aps ] ) , up to the replacements @xmath48 and @xmath49 in eq . ( [ spec - aps ] ) . in the present paper , we revisit induced coherent radiation and consider all @xmath0 hard forward processes , namely @xmath1 , @xmath2 , @xmath3 and @xmath4 , assuming the hard exchange in the @xmath35-channel to be color octet ( for @xmath1 and @xmath2 ) or color triplet ( for @xmath3 and @xmath4 ) . we work in the same setup as ref . @xcite , but derive the induced coherent spectrum in full rigor , to all orders in the opacity expansion and for the specific case of a coulomb rescattering potential . as a main result , we find that the induced coherent spectrum associated to @xmath32 hard forward processes is given by the general ( approximate ) expression ( [ spec - general ] ) ( or equivalently ( [ spec - general - pavspp ] ) ) , with @xmath50 , @xmath51 the color charges of the incoming and outgoing partons , @xmath52 the color charge of the @xmath35-channel exchange , and @xmath53 the _ typical _ transverse momentum broadening induced by multiple coulomb rescatterings . the results obtained previously for the spectra associated to @xmath34 @xcite and @xmath2 @xcite are simply interpreted as particular cases of eq . ( [ spec - general ] ) , corresponding respectively to @xmath54 and @xmath55 . our paper is organized as follows . first , the case of an asymptotic parton of color charge @xmath50 scattering at small angle off a _ pointlike _ target through single gluon exchange is studied in section [ sec : pointlike ] . the radiation spectrum associated to this process is determined in the soft radiation ( @xmath56 ) and high - energy limit , see eq . ( [ gbspec - exact ] ) . this allows one to recover the gunion - bertsch spectrum , eq . ( [ gbspec ] ) , in the eikonal limit ( scattering angle @xmath57 ) as well as the spectrum in the hard scattering limit ( @xmath58 ) , eq . ( [ gbspec - hard ] ) . the @xmath59-integrated spectrum @xmath60 is discussed in section [ sec : qdep ] , where we note that the _ variation _ of the spectrum with respect to the hard exchange @xmath61 arises from gluon emission in the ` soft ' ( @xmath62 ) and ` hard ' ( @xmath63 ) transverse momentum regions . we argue that the color factor of the soft contribution heralds that of the induced spectrum in a finite size target . the gluon spectrum associated to the scattering of an asymptotic parton off a _ finite - size _ target is investigated in detail in section [ sec : hard - smallangle ] , for the case of a _ coulomb _ rescattering potential . the medium - induced gluon spectrum is determined rigorously in the soft ( @xmath64 ) and large formation time ( @xmath27 ) limits . calculations are carried out at first order in the opacity expansion , @xmath65 , eq . ( [ specn1 ] ) , and to all orders , @xmath66 , eqs . ( [ spec - alln])([sar ] ) . we observe that the radiation spectrum is given by the overlap between the initial parton gluon wavefunction ( ` evolved ' by soft rescatterings , and thus medium dependent ) and the wavefunction of the outgoing parton gluon system . when @xmath66 the induced spectrum can be approximated by a simple expression , eq . ( [ spec - alln - appr ] ) , which in the case @xmath67 coincides with eq . ( [ spec - aps-0 ] ) derived semi - heuristically in @xcite , provided @xmath68 in ( [ spec - aps-0 ] ) is understood as the _ typical _ exchange in multiple coulomb scattering , @xmath69 . the results obtained in section [ sec : hard - smallangle ] for the scattering of a fast asymptotic parton ( _ i.e. _ @xmath70 or @xmath71 ) are generalized to other processes in section [ sec : other - processes ] . the case of an incoming fast quark scattered to an outgoing fast gluon ( @xmath72 ) is first examined . we find that the associated medium - induced spectrum is identical to that obtained in the case of an asymptotic gluon ( @xmath71 ) , despite the different color charge in the initial state . the case of a fast gluon scattered to a pointlike color octet of mass @xmath37 ( like for instance a compact heavy @xmath36 pair ) is also treated . as for @xmath71 , the medium - induced gluon spectrum associated to @xmath34 can be accurately approximated by a simple expression ( eq . ( [ massive - spec - appr ] ) to all orders in the opacity expansion ) , which has been used for phenomenological applications in @xcite . we also discuss in section [ sec : is - fs ] the purely initial - state ( respectively , final - state ) radiation , associated to hard processes where the color charge of the outgoing particle ( respectively , incoming particle ) vanishes and coherent radiation between initial and final states is thus absent . in those cases , we explicitly verify that radiation with @xmath27 does not contribute to the induced gluon spectrum . the purely initial - state ( or final - state ) induced spectrum @xmath60 arises from radiation with @xmath73 , leading to the known result @xmath26 , independent ( up to logarithms ) of the parton energy . finally , we summarize our main results in section [ sec : discussion ] , where we state and discuss the general ( approximate ) formulae ( [ spec - general ] ) ( or equivalently ( [ spec - general - pavspp ] ) ) for the induced coherent spectrum associated to hard forward @xmath0 processes , at all orders in the opacity expansion . in this final section we also discuss the connection of our study with a previous study of gluon production in the saturation formalism @xcite . here we briefly review the radiation spectrum associated to small angle scattering of an asymptotic parton off a pointlike target . consider a massless quark of large momentum @xmath74 , with @xmath75 and @xmath76 . ] acquiring the transverse momentum @xmath77 via single gluon exchange , see fig . [ fig : gbamplitude]a . we focus on small angle scattering , @xmath78 . the gluon emission amplitude induced by the latter scattering is given by the diagrams of fig . [ fig : gbamplitude]b , and was first derived by gunion and bertsch @xcite . denoting the radiated gluon momentum @xmath79 and focussing on soft ( @xmath56 ) and small angle ( @xmath80 ) radiation , the gauge invariant radiation amplitude reads @xcite [ gb ] m_rad = _ el ( 2 g ) . in this expression @xmath81 is the qcd coupling , @xmath82 stands for the ( real ) _ physical _ polarization of the radiated gluon , formally disappears after squaring the amplitude and summing over the two physical polarization states , it will be implicit in the following . ] and @xmath83 , @xmath84 , @xmath85 are the _ full _ color factors associated to the diagrams ( i ) , ( ii ) , ( iii ) of fig . [ fig : gbamplitude]b . hence @xmath86 corresponds to the lorentz part of the elastic amplitude @xmath87 of fig . [ fig : gbamplitude]a , _ i.e. _ , @xmath88 with @xmath89 the ` elastic ' color factor . in light - cone perturbation theory and in light - cone gauge @xmath90 , it is easy to check that the first two terms in the r.h.s . of ( [ gb ] ) correspond to initial state radiation ( diagrams ( i ) and ( ii ) of fig . [ fig : gbamplitude]b ) , and the last term to final state radiation ( diagram ( iii ) ) . the gunion - bertsch amplitude is then very simply interpreted , since @xmath91 is nothing but the light - cone wavefunction ( in light - cone gauge ) of the quark - gluon fluctuation in the incoming quark state @xcite . the radiated gluon with @xmath92 and @xmath93 can thus arise directly from the incoming quark - gluon fluctuation @xmath94 ( diagram ( i ) ) , from the incoming fluctuation @xmath95 followed by gluon scattering ( diagram ( ii ) ) , or from the quark - gluon fluctuation in the quark _ after _ the scattering ( diagram ( iii ) ) . the latter fluctuation has a total transverse momentum @xmath61 and thus a wavefunction @xmath96 , as dictated by lorentz transformation laws of light - cone wavefunctions @xcite . of an energetic parton off a pointlike target . ( b ) induced gluon radiation amplitude @xmath97.,width=453 ] the general structure of the gunion - bertsch amplitude ( [ gb ] ) is actually independent of the incoming parton type , as well as of the target type . in the following we consider an energetic parton of color charge @xmath50 , with @xmath98 for a gluon and @xmath99 for a quark . the gluon radiation intensity @xmath100 is obtained from ( [ gb ] ) as di = = ( ) , [ phase - space - factor ] where @xmath101 and @xmath102 are summed over initial and final color indices . this gives the radiation spectrum [ qcd - spec - pict ] x = , where we use a pictorial representation defined as follows . for each diagram appearing in the numerator of ( [ qcd - spec - pict ] ) , the upper ( lower ) part arises from one of the three contributions to the emission amplitude ( conjugate amplitude ) given in ( [ gb ] ) or in fig . [ fig : gbamplitude]b . the lorentz part of each diagram follows from the rules depicted in fig . [ fig : pictorial - rules]a , and the color factor from the rules of fig . [ fig : pictorial - rules]b . the diagram in the denominator of ( [ qcd - spec - pict ] ) is simply the color factor associated to the elastic cross section . we can thus rewrite ( [ qcd - spec - pict ] ) explicitly as x = \ { c_r \| \|^2 + c_r \| \|^2 + n_c \| \|^2 . 10 mm & & + . - 2 -2 ( c_r - ) - 2 } . 0 mm & & grouping together the terms @xmath103 and those @xmath104 we obtain [ gbspec - exact ] x = \ { c_r + n_c } . the latter expression for the radiation spectrum induced by a single transverse momentum exchange @xmath61 is exact ( within the soft radiation and high - energy limit defined above ) , and will be used in the following to review several limiting cases . is denoted by the solid line.,width=453 ] in ( [ gbspec - exact ] ) the term @xmath103 is the abelian - like contribution to the spectrum . indeed , in qed color factors as well as the diagram ( ii ) of fig . [ fig : gbamplitude]b are absent , and the photon emission amplitude off an electron is of the form [ qed - amp ] m_rad^qed e , which square reproduces the first term of ( [ gbspec - exact ] ) . this amplitude vanishes in the limit @xmath105 , _ i.e. _ , when the scattering angle of the incoming electron @xmath106 is negligible compared to the photon emission angle @xmath107 . in qed , a non - zero scattering angle @xmath108 is necessary to perturb the electron quantum state and thus to induce radiation . as a consequence , the abelian - like contribution to the @xmath59-integrated spectrum @xmath109 arises from @xmath110 . explicitly , [ qed - spec ] . x @xmath111 to regularize the singularities at @xmath112 and @xmath113.the integral in ( [ qed - spec ] ) is regularized by replacing @xmath114 and @xmath115 in the denominator of the integrand . taking then the limit @xmath116 yields the r.h.s . of ( [ qed - spec ] ) . ] those singularities contribute equally to ( [ qed - spec ] ) , the logarithm arising either from the domain @xmath117 ( singularity at @xmath112 ) , or from the domain @xmath118 ( singularity at @xmath113 ) . this corresponds to the two well - known cones , centered at @xmath112 and @xmath113 and of opening @xmath108 , of abelian - like radiation . in qcd , gluon radiation can occur even in the eikonal limit where the energetic parton trajectory is approximated as a straight line ( @xmath57 ) , due to the parton _ color rotation _ in the elastic scattering . indeed , in the limit @xmath105 the amplitude ( [ gb ] ) remains finite , [ qcd - amp ] m_rad g , where @xmath119 are color matrices in the representation of the parton , @xmath120 . the eikonal limit allows to single out the purely non - abelian contribution to the radiation amplitude , as reflected by the commutator in ( [ qcd - amp ] ) . the resulting spectrum , obtained by neglecting @xmath121 compared to @xmath59 in ( [ gbspec - exact ] ) , is the so - called gunion - bertsch spectrum @xcite in the same @xmath59 range , [ gbspec ] x ( x q_k _ ) . we now consider the ` hard scattering limit ' , defined as @xmath122 , but still keeping a small scattering angle , @xmath123 . in this limit the second term of ( [ gb ] ) ( diagram ( ii ) of fig . [ fig : gbamplitude]b ) can be neglected , leading to [ gbhard ] m_rad g . the associated spectrum can be read off from ( [ gbspec - exact ] ) , [ gbspec - hard ] x \ { c_r + n_c } ( k_q _ ) . in addition to the two narrow ` abelian cones ' of opening @xmath108 ( term @xmath103 ) , there is a non - abelian term @xmath104 ) the non - abelian contribution is @xmath104 due to our assumption of a single gluon exchange . in general the purely non - abelian contribution is proportional to the color charge exchanged in the @xmath35-channel of the elastic scattering @xcite . the same remark applies to the gunion - bertsch spectrum ( [ gbspec ] ) . ] arising from emission angles _ larger _ than the scattering angle @xmath108 @xcite . the latter property follows from averaging the non - abelian contribution over the azimutal angle of @xmath124 using the identity [ azim - integral ] = ( k^2 - a^2 ) . effectively , ( [ gbspec - hard ] ) can thus be written as [ gbspec - hard-2 ] x \ { c_r + n_c } ( k_q _ ) . when @xmath125 , the abelian - like and non - abelian contributions have similar magnitudes . in the region @xmath126 , the non - abelian contribution dominates , and the spectrum ( [ gbspec - hard-2 ] ) obviously coincides with the gunion - bertsch spectrum ( [ gbspec ] ) , [ log - interval ] x ( x q_k_q _ ) . in situations where only the final energetic parton is tagged , the relevant quantity is the _ total _ , @xmath59-integrated spectrum @xmath127 . strictly speaking , the integral over @xmath59 of ( [ gbspec - exact ] ) is ill - defined due to singularities at @xmath112 , @xmath113 and @xmath128 , and is not a physical observable . in particular it includes collinear radiation with respect to the initial ( @xmath112 ) and final ( @xmath113 ) parton directions . however , what usually matters in practice is the _ variation _ of the total spectrum with @xmath61 , which variation is infrared safe and physical . here we discuss the total radiation spectrum associated to a single scattering @xmath61 , and the variation of the total spectrum with @xmath61 . the main goal of this discussion is to provide some insight to the medium - induced spectrum in a finite size target studied in section [ sec : hard - smallangle ] . in particular , we will show in section [ sec : hard - smallangle ] that the induced spectrum off a parton of color charge @xmath50 has a color factor @xmath129 , _ @xmath130 for an energetic gluon , and @xmath131 for a quark . a _ negative _ induced energy loss of a quark is somewhat unusual , but can be intuitively understood from the case of a pointlike target . indeed , we show below that the factor @xmath129 is precisely the color factor of the contribution to the total spectrum arising from the ` soft ' @xmath132-domain , which turns out to be the only domain contributing to the induced loss . in the following we regularize the @xmath59-integral by introducing an infrared cutoff @xmath111 , as was done in section [ sec : abelian ] for the abelian - like contribution ( see footnote [ footreg ] ) . the same results can be obtained using dimensional regularization . the exact @xmath59-differential spectrum ( [ gbspec - exact ] ) being approximated by ( [ gbspec ] ) for @xmath105 and by ( [ gbspec - hard-2 ] ) for @xmath133 , the total spectrum can be simply obtained to logarithmic accuracy by introducing an arbitrary separation scale @xmath134 satisfying @xmath135 . the contribution to the total spectrum arising from @xmath136 is obtained from ( [ qed - spec ] ) and ( [ gbspec - hard-2 ] ) , , the abelian - like contribution ( [ qed - spec ] ) is modified by terms @xmath137 which are negligible compared to the dominant logarithmic term . similarly , since ( [ gbspec - hard-2 ] ) is an approximation at @xmath133 , there are corrections @xmath138 ( which can also be dropped ) to the term @xmath104 in eq . ( [ soft - k ] ) . ] [ soft - k ] . the contribution from @xmath139 follows from ( [ gbspec ] ) , [ hard - k ] . x \|_k__s ( k^2 - _ s^2 ) , where the singularity at @xmath128 is screened by the infrared cutoff . to logarithmic accuracy , the integral in ( [ hard - k ] ) is dominated by two logarithmic intervals , namely @xmath140 and @xmath141 , and thus reads ) can be calculated analytically for any fixed @xmath142 , @xmath134 and @xmath111 . taking then the limit @xmath116 , we obtain ( [ hard - k-2 ] ) up to corrections @xmath138 . ] [ hard - k-2 ] . x \|_k__s . when adding the contributions ( [ soft - k ] ) and ( [ hard - k-2 ] ) , the dependence on the arbitrary scale @xmath134 cancels out , as expected . second , as already mentioned , the total spectrum diverges in the limit @xmath116 but its variation when @xmath142 is shifted to @xmath143 is infrared - safe , [ spectrum - shift ] . x \|_q , where the first and second terms arise respectively from the ` soft ' ( @xmath144 ) and ` hard ' ( @xmath145 ) domains . the modification of the soft contribution ( @xmath146 ) with respect to an increase of @xmath142 is proportional to @xmath147 . indeed , in the soft part ( [ soft - k ] ) the abelian - like radiation at angles smaller than the scattering angle ( @xmath148 , first term of ( [ soft - k ] ) ) _ increases _ with a rate @xmath149 , but the purely non - abelian radiation arising from larger angles ( @xmath150 , second term of ( [ soft - k ] ) ) , _ decreases _ with a rate @xmath104 . hence a net variation @xmath151 . in the particular case of an asymptotic quark , we have @xmath152 , implying that the number of gluons ( of a given energy ) radiated into a cone of fixed size _ decreases _ with increasing momentum transfer from the target . at the same time , the _ total _ number of radiated gluons increases with a rate @xmath153 , due to radiation of gluons with @xmath63 . although the spectrum at zeroth order in opacity receives contributions from both soft and hard @xmath132-domains , we will see in section [ sec : hard - smallangle ] that in the _ medium - induced _ spectrum , the gluon @xmath132 is constrained to be _ soft _ , the additional scale @xmath154 coming into play in a finite size target acting as an upper cut - off . we thus expect the induced radiation to correspond to the medium - modification of the _ soft part _ of the total spectrum , and to have the same color factor @xmath155 . we now consider the case where the energetic parton crosses a nuclear target of thickness @xmath6 , still assuming a small deviation angle @xmath156 and a negligible longitudinal momentum transfer to the medium . the parton can undergo several _ soft _ scatterings @xmath157 in the target , leading to nuclear transverse momentum broadening @xmath158 , with @xmath159 . we focus on the kinematics where the final parton is ` tagged ' with a transverse momentum @xmath160 much larger than nuclear broadening . thus @xmath160 must arise dominantly from a _ single hard _ scattering @xmath161 . we want to derive the medium - induced gluon radiation spectrum associated to such a hard scattering accompanied by any number of soft rescatterings . we focus on soft radiation as compared to the hard process , _ i.e. _ , @xmath162 but also @xmath163 . the latter assumption will be justified a posteriori , when verifying that the @xmath59-integrated medium - induced spectrum arises dominantly from @xmath164 . ) and the discussion after ( [ elossn1 ] ) . for the calculation at all orders in opacity , see the final comments of section [ useful - appr ] . ] the medium - induced radiation spectrum can be derived in an opacity expansion @xcite . at zeroth order , _ i.e. _ , in absence of soft rescattering , the spectrum associated to the hard scattering @xmath165 is given by ( [ gbspec - hard ] ) and cancels in the induced radiation spectrum . at a generic order @xmath166 , the latter reads for a simple derivation of ( [ spec - order - n ] ) in the limit of large formation time @xmath27 . ] x = d^2 k [ spec - order - n ] using the pictorial rules defined in section [ sec : pointlike ] ( see fig . [ fig : pictorial - rules ] ) . the ` elastic ' color factor @xmath167 reads [ fn ] f_n = = d_r ( c_r)^n+1 , where @xmath168 is the dimension of the parton color representation , @xmath169 for a quark and @xmath170 for a gluon . in ( [ spec - order - n ] ) @xmath171 is the parton elastic mean free path and @xmath172 is the probability distribution for the transverse momentum transfer in the elastic scattering . modelling each scattering center as a static center creating a screened coulomb potential @xcite , we have [ coulomb - pot ] v ( ) = ; d^2 v ( ) = 1 . as in ref . @xcite , we choose @xmath171 and the screening length @xmath173 of the scattering potential as the model parameters and assume @xmath174 , implying that successive soft scatterings can be treated as independent . they can thus be ordered in light - cone time @xmath175 , which in the high - energy limit is equivalent to ordering in usual time @xmath35 or longitudinal position , _ i.e. _ , @xmath176 in ( [ spec - order - n ] ) . the longitudinal position of the hard scattering @xmath165 will be denoted as @xmath177 , with @xmath178 . the sum of diagrams in ( [ spec - order - n ] ) is in general complicated to evaluate . however , anticipating that the medium - induced radiation spectrum @xmath60 is dominated , in the present case of small angle scattering , by large gluon formation times @xmath27 , the calculation can be greatly simplified by working in this limit from the beginning . when @xmath27 , the dominant diagrams are those where the gluon emission times in the radiation amplitude and conjugate amplitude , denoted respectively by @xmath35 and @xmath179 , occur either before or after the nuclear target . the diagrams where @xmath35 ( or @xmath179 ) occurs within the target , @xmath180 , are proportional to the difference @xmath181 between two ` phase factors ' , where @xmath182 ( see ref . @xcite and section [ sec : is - fs ] below ) , and vanish when @xmath27 . the dominant diagrams in the sum of ( [ spec - order - n ] ) can thus be classified into three classes , namely ( i ) @xmath183 ( final state radiation ) ( ii ) @xmath184 ( initial state radiation ) and ( iii ) @xmath185 and @xmath186 ( interference ) . ) belongs to the latter class . ] another simplification arises from our assumption @xmath122 . in this limit the diagrams where the hard gluon @xmath61 connects to the radiated gluon are negligible , similarly to the case of a pointlike target ( section [ sec : hard - scatt - pointlike ] ) . finally , let us stress that in the limit @xmath187 , the radiation can not probe the target structure , but this does not imply that the radiation spectrum is medium - independent . indeed , the spectrum may depend on the target size @xmath6 through the amount of soft rescattering @xmath188 ( or through the _ probability _ for such rescattering ) suffered by the incoming parton - gluon fluctuation across the target . this results in a non - vanishing medium - induced radiation , as confirmed in the following sections by an explicit calculation , at first and all orders in the opacity expansion . at order @xmath189 , and within the limit @xmath27 and @xmath122 , the spectrum ( [ spec - order - n ] ) is of the form x = d^2 k d^2 _ 1 v(_1 ) , [ spec - order-1 ] where the sets of diagrams @xmath190 , @xmath191 and @xmath192 are shown in fig . [ fig : abcn1 ] . when @xmath27 one may approximate @xmath193 in those diagrams , which thus become independent of @xmath177 and @xmath194 and in particular of their precise ordering . as a consequence we have @xmath195 in ( [ spec - order-1 ] ) . , @xmath191 , @xmath192 of diagrams , at first order in the opacity expansion , which may contribute to the radiation spectrum in the limit @xmath187 , see ( [ spec - order-1 ] ) . only set @xmath192 turns out to be non - vanishing.,width=574 ] \(i ) set @xmath190 : @xmath183 the set @xmath190 of diagrams corresponds to the class ( i ) defined in section [ sec : setup ] , namely @xmath183 ( final state radiation ) , and is obtained from the rules of fig . [ fig : pictorial - rules ] as [ setan1 ] d^2 k = d^2 k \ { c_r \| \|^2 + 2 ( -1 ) c_r \| \|^2 } = 0 . the virtual contribution ( right diagram of fig . [ fig : abcn1]a ) exactly cancels the real one ( left diagram of fig . [ fig : abcn1]a ) . the relative weights of virtual corrections have been discussed previously , see for instance ref . they contribute a factor @xmath196 , and a symmetry factor @xmath197 when the two gluon lines ( @xmath157 and @xmath198 ) connect to the same parton line ( energetic parton or radiated gluon ) . the associated color factor is obtained from the rules of fig . [ fig : pictorial - rules]b . the vanishing of the set @xmath190 can be understood as follows . when @xmath183 , the gluon is effectively produced after the nuclear medium . gluon radiation should thus be similar to that in vacuum , up to the change in the direction of the final parton due to the rescattering @xmath199 . when integrating over @xmath59 , the shift in the final parton direction becomes irrelevant , and the contribution to the _ medium - induced _ radiation spectrum vanishes . considered here , which amounts to neglect the deviation angle of the energetic parton due to soft rescatterings , the vanishing of the set @xmath190 occurs at fixed @xmath59 , _ i.e. _ , at the integrand level in ( [ setan1 ] ) . ] in general we expect no contribution to the @xmath59-integrated , medium - induced spectrum when @xmath35 and @xmath179 belong to a same time - interval where no rescattering happens . \(ii ) set @xmath191 : @xmath184 we now turn to the set @xmath191 of diagrams , see fig . [ fig : abcn1]b , corresponding to the class ( ii ) @xmath184 ( initial state radiation ) . applying the rules of fig . [ fig : pictorial - rules ] , the sum of the six diagrams of the set @xmath191 simplifies to [ setbn1 ] d^2 k = n_c d^2 k = 0 . the general property mentioned above is at work . when @xmath184 , the gluon is effectively radiated before the nuclear target , and the following soft rescatterings can not affect the total ( @xmath59-integrated ) radiation rate . \(iii ) set @xmath192 : @xmath200 and @xmath201 in the limit @xmath27 , the only non - vanishing contribution to @xmath60 arises from the set @xmath192 of interference diagrams with @xmath200 and @xmath201 , see fig . [ fig : abcn1]c . applying the diagrammatic rules we obtain ( @xmath202 ) [ setcn1 ] d^2 k = d^2 k \ { c_r . & & + + . } . & & when @xmath203 , the latter expression simplifies and the medium - induced spectrum ( [ spec - order-1 ] ) becomes ( use @xmath204 ) [ spectrum - n1 ] x = ( 2 c_r - n_c ) d^2 kd^2 v ( ) . the last factor in the integrand of ( [ spectrum - n1 ] ) is proportional to the ( conjugate ) wavefunction @xmath205 of the final parton - gluon fluctuation , see section [ sec : exact - pointlike ] . the other factor in between brackets can be interpreted as the incoming ` medium - induced wavefunction ' . indeed , this factor is given by the incoming wavefunction ` evolved ' by rescatterings @xmath206 , from which the ` vacuum ' wavefunction @xmath207 is subtracted . this interpretation of the radiation spectrum as the overlap between initial and final wavefunctions of the parton - gluon fluctuation holds to all orders in opacity , as we will see in section [ sec : allorders ] , see ( [ spec - alln - rescaled ] ) . the expression ( [ spectrum - n1 ] ) can be simplified by averaging over the azimutal angles of @xmath208 and @xmath61 . using ( [ azim - integral ] ) we get [ kbounds ] = , leading to [ ipp ] x = ( 2 c_r - n_c ) _ x^2 ^2^ d^2 . the integral over @xmath188 can be performed using a simple integration by parts , giving the final expression for the medium - induced spectrum at first order in the opacity expansion , [ specn1 ] x = ( 2 c_r - n_c ) . the associated average energy loss is proportional to @xmath24 , [ elossn1 ] p^+ p^+ dx x = ( 2 c_r - n_c ) _ s p^+ . due to the fast decrease of the spectrum ( [ specn1 ] ) as @xmath209 when @xmath210 , the energy loss ( [ elossn1 ] ) is dominated by @xmath211 . using ( [ kbounds ] ) and ( [ ipp ] ) we infer @xmath212 , and the energy loss is thus dominated by gluon formation times t_~ ~ which scale as @xmath24 when @xmath213 . this justifies our working assumption of large formation times @xmath27 . the radiative energy loss ( [ elossn1 ] ) arises from gluon radiation which is _ fully coherent _ over the medium size @xmath6 . the overall color factor @xmath147 of the energy loss ( [ elossn1 ] ) equals @xmath130 for an energetic gluon , and @xmath131 for a quark . the latter suppression in the large @xmath130 limit is simply understood , since in the quark case the relevant diagrams ( fig . [ fig : abcn1]c ) are non - planar @xcite . the fact that the _ quark _ induced loss is _ negative _ is reminiscent from the case of a pointlike target , see the discussion in section [ sec : qdep ] . we have seen there that the total radiation spectrum induced by a single hard scattering @xmath61 acquires its @xmath61-dependence from two domains of @xmath132 , the soft @xmath214 and hard @xmath215 domains , see ( [ spectrum - shift ] ) , with associated color factors @xmath155 and @xmath216 respectively . in the case of _ medium - induced _ radiation studied here , the rescattering @xmath208 effectively provides an upper cut - off on the radiated gluon transverse momentum , @xmath217 , see ( [ kbounds ] ) . this justifies the dominance of soft @xmath132 for medium - induced radiation , and elucidates the origin of the overall color factor . it is instructive to state the result for the medium - induced radiation spectrum in qed . replacing @xmath218 , @xmath219 in ( [ setcn1 ] ) and substituting it into ( [ spec - order-1 ] ) we obtain [ specn1-qed ] . x \|_qed = _ & & + = _ , & & with @xmath220 the electron mean free path and @xmath221 . the interpretation of the latter spectrum as a _ medium - induced _ spectrum is clear from the second line of ( [ specn1-qed ] ) , which is the difference between abelian - like spectra ( [ qed - spec ] ) for single effective scatterings @xmath222 and @xmath61 respectively . in qed the medium - induced spectrum vanishes when the electron scattering angle is the same with or without rescattering , @xmath223 . in ref . @xcite , the medium - induced spectrum in qed was derived assuming _ both _ @xmath27 _ and _ @xmath224 ( in our notations ) , and was found to vanish . this led the authors to conclude that the assumption @xmath27 is invalid , leaving only radiation with small formation time @xmath225 , resulting in some bound on induced energy loss , @xmath226 when @xmath227 . we stress here that it is instead the approximation @xmath224 which is too drastic . indeed , relaxing the latter and keeping @xmath27 , a non - zero medium - induced spectrum follows . shifting @xmath228 in the first line of ( [ specn1-qed ] ) , we see that the spectrum is actually independent of @xmath229 , and thus given by the same expression with @xmath229 set to unity . changing then @xmath230 , the integral over @xmath59 becomes identical to the integral in ( [ spectrum - n1 ] ) evaluated at @xmath231 . we can thus read off the result from ( [ specn1 ] ) , . x \|_qed = , leading to the medium - induced energy loss ( use @xmath232 ) [ elossqed ] . p^+ \|_qed p^+ . to summarize , the brodsky - hoyer bound does not apply to the case under study of small angle scattering of an energetic charge , neither in qed ( see ( [ elossqed ] ) ) nor in qcd ( see ( [ elossn1 ] ) ) . in both cases the scaling @xmath23 originates from radiation with large formation times @xmath27 , _ i.e. _ , which is fully coherent over the medium size . in qed the medium - induced spectrum arises from the electron scattering angle being affected by in - medium rescatterings ( @xmath233 ) , whereas in qcd a non - vanishing spectrum is found even in the limit @xmath234 , due to the parton color rotation . finally , let us note that the brodsky - hoyer argument does apply to purely initial or purely final state radiation , both in qed and qcd . indeed , in those cases radiation with large formation time cancels out , see ( [ setan1 ] ) and ( [ setbn1 ] ) . as we will review in section [ sec : is - fs ] , this leads to a medium - induced energy loss satisfying ( up to logarithms , see ( [ fs - eloss ] ) ) the brodsky - hoyer bound . here we derive the radiation spectrum to all orders in the opacity expansion . at any order @xmath166 , it is clear that in the limit @xmath27 the dominant diagrams are the interference diagrams corresponding to @xmath200 and @xmath201 , as in the case @xmath189 , see fig . [ fig : abcn1]c . it is convenient to rewrite ( [ spec - order - n ] ) as x = d^2 k f_n(k , l ) [ spec - order - n - bis ] f_n(k , l ) ( ) ^n , [ fn ] where the quantity @xmath235 corresponds to the set of diagrams of fig . [ fig : cnplus1]a , defined by the rules of fig . [ fig : pictorial - rules ] but with the emission factor in the conjugate amplitude @xmath236 _ removed_. contributing to ( [ fn ] ) . the shaded area denotes all possible attachments of the lines @xmath237 , and the cross in the conjugate amplitude means that the emission vertex @xmath236 is not included in the definition of @xmath235 . ( b ) diagrammatic expression of @xmath238 in terms of @xmath235 , leading to ( [ c - rec]).,width=529 ] the medium - induced spectrum to all orders in the opacity expansion is obtained by summing ( [ spec - order - n - bis ] ) over @xmath166 , & & x = _ n=1^ x = d^2 k , [ spec - alln-0 ] + & & f(k , l ) _ n=0^ f_n(k , l ) ; f_0(k , l ) f_0(k ) = ( 2c_r - n_c ) . [ f - def ] the function @xmath239 can be derived by noticing that at order @xmath240 , the set of diagrams @xmath238 is given by the recurrency relation ( see fig . [ fig : cnplus1]b ) c_n+1(k ; \ { _ 1 _ n+1 } ) = -n_c c_n(k ; \ { _ 1 _ n } ) + n_c c_n(k- _ n+1 ; \ { _ 1 _ n } ) , [ c - rec ] where the first term arises from the first four diagrams , and the second term from the last two diagrams of fig . [ fig : cnplus1]b . from ( [ fn ] ) and ( [ c - rec ] ) we obtain = - d^2 v ( ) . [ rec - rel ] using @xmath204 and defining @xmath241 , we easily show that @xmath239 defined in ( [ f - def ] ) satisfies the equation = - f(k , r ) + d^2 v ( ) f(k- , r ) ; f(k , r=0 ) = f_0(k ) . [ equation - f ] rescaling @xmath242 by the factor @xmath243 , the spectrum ( [ spec - alln-0 ] ) becomes x = ( 2c_r - n_c ) d^2 k ( - f_0(k- xq ) ) , [ spec - alln - rescaled ] where @xmath242 is solution of ( [ equation - f ] ) with the rescaled initial condition @xmath244 . we easily check that keeping in ( [ spec - alln - rescaled ] ) only the @xmath189 term of the opacity expansion , we recover the result ( [ spectrum - n1 ] ) obtained in section [ sec : n1qcd ] . in the bracket of ( [ spec - alln - rescaled ] ) , and get @xmath245 from ( [ rec - rel ] ) . ] as is explicit from ( [ spec - alln - rescaled ] ) , the spectrum is given by the overlap between the ( conjugate ) wavefunction @xmath246 of the final state parton - gluon fluctuation and the ` medium - induced wavefunction ' @xmath247 , where @xmath242 is obtained by evolving the incoming wavefunction @xmath248 according to ( [ equation - f ] ) . the latter equation accounts for the modification of the radiated gluon transverse momentum @xmath59 due to multiple soft rescatterings , and is identical to the equation satisfied by the gluon transverse momentum probability distribution @xmath249 after a ` time ' @xmath250 travelled in the medium , see ref . @xcite . by going to the impact parameter space , ( b , r ) = d^2 k f(k , r ) e^- i kb , [ impact - space ] the set of equations ( [ equation - f ] ) and ( [ spec - alln - rescaled ] ) becomes & & = - ( 1 - v(b ) ) ( b , r ) ; ( b , r=0 ) = _ 0(b ) = - 2i , [ equation - f - bspace ] + & & x = ( 2c_r - n_c ) [ ( b , r ) - _ 0(b ) ] ( - _ 0(-b ) ) e^i x qb . [ spec - alln - bspace ] the solution of ( [ equation - f - bspace ] ) is ( b , r ) = _ 0(b ) , and ( [ spec - alln - bspace ] ) can be rewritten as x = ( 2c_r - n_c ) \ { 1 - } e^i x qb . [ spec-5 ] finally , using @xmath251 for the coulomb potential ( [ coulomb - pot ] ) , performing the azimutal integral in ( [ spec-5 ] ) and changing variable @xmath252 , we obtain x & = & ( 2c_r - n_c ) s [ ; r ] ; ; r , [ spec - alln ] + s[;r ] & & 2 _ 0^ j_0(b ) \ { 1 - } . [ sar ] the medium - induced spectrum to all orders in the opacity expansion given by eqs . ( [ spec - alln])([sar ] ) is one of the main results of our study . let us emphasize that the factor @xmath253 in the exponential s argument is ( up to a normalization factor ) the _ dipole _ scattering cross section of a color singlet @xmath254 pair on a color center , see for instance ref . @xcite . here we focus on the parametric dependence of the exact spectrum ( [ spec - alln ] ) ( _ i.e. _ of the function @xmath255 $ ] defined in ( [ sar ] ) ) in various limits . for simplicity we however assume a medium of large size , @xmath256 , and derive the limits _ at fixed @xmath66 _ of @xmath255 $ ] at small and large @xmath257 . first , we note that at large @xmath256 , the integral over @xmath191 in ( [ sar ] ) is dominated by small @xmath258 , and we can thus approximate @xmath259 at small @xmath191 , giving [ sar - appr ] s[;r ] _ r 1 _ 0 ^ 1 j_0(b ) \ { 1 - } . in order to extract the small @xmath257 limit of the latter expression ( the precise meaning of ` small @xmath257 ' to be defined shortly ) , we first note that when @xmath260 , the integral ( [ sar - appr ] ) is dominated by @xmath261 . to _ logarithmic accuracy _ ( @xmath262 ) , we can thus replace @xmath263 in the exponential of ( [ sar - appr ] ) , and after the change of variable @xmath264 we obtain s[;r ] _ 0^ j_0 ( ) ( 1 - e^-u/4 ) = -ei ( - ) , [ u - int ] where we have set @xmath265 in the upper bound of the @xmath266-integral , and @xmath267 . thus , at ` small @xmath257 ' , @xmath268 $ ] has an approximate scaling with the variable @xmath269 , as illustrated in fig . [ fig : appr - scaling ] ( left ) . this very fact implies that the ` small @xmath257 ' limit should be defined as @xmath270 . since ( [ u - int ] ) strictly holds only in this limit , we can use , without loss of generality , @xmath271 in the r.h.s . of ( [ u - int ] ) , leading to [ s - log - int ] s[;r ] when . this yields the limiting behavior of the spectrum ( [ spec - alln ] ) at small @xmath229 , [ small - x ] x ( 2c_r - n_c ) when x . we stress that in order to obtain the small @xmath229 limit , we have replaced @xmath263 in the exponential of ( [ sar - appr ] ) , which amounts to neglect the logarithmic dependence on @xmath272 of the dipole scattering cross section at small @xmath272 , and thus to approximate the latter as an exactly quadratic function of @xmath272 , namely @xmath273 . in other words , the so - called _ harmonic oscillator _ approximation ( see _ e.g. _ @xcite ) allows one to access the correct small @xmath229 limit of the coherent spectrum . we also note that our initial working assumption @xmath163 was legitimate in order to derive ( [ small - x ] ) . indeed , from the above discussion ( following ( [ sar - appr ] ) ) , the behavior ( [ small - x ] ) arises from impact parameters @xmath274 , _ i.e. _ , from gluon transverse momenta @xmath275 , with @xmath203 within the setup defined in section [ sec : setup ] . when @xmath276 visible deviations to the scaling of @xmath268 $ ] in @xmath277 appear , see fig . [ fig : appr - scaling ] ( left ) . indeed , at large @xmath257 the above derivation does not hold , due to the rapid oscillations of @xmath278 in ( [ sar - appr ] ) . at large @xmath257 , we show in appendix [ app : limits ] that @xmath268 $ ] scales with @xmath279 ( rather than with @xmath280 ) and behaves as s[;r ] when . [ large - omega ] this gives the limiting behavior of the coherent spectrum ( [ spec - alln ] ) at large @xmath229 , [ large - x ] x ( 2c_r - n_c ) when x . it can be easily verified ( see appendix [ app : limits ] ) that obtaining the large @xmath229 behavior ( [ large - x ] ) requires keeping the exact @xmath272-dependence of the dipole scattering cross section at small @xmath272 , namely @xmath281 . this can be simply understood as follows . at large @xmath282 , the integral ( [ sar - appr ] ) is not dominated by @xmath261 any longer , but by smaller values @xmath283 , hence the importance of keeping the @xmath284 factor in the exponential when @xmath285 . the harmonic oscillator approximation would not be adequate to derive the parametric behavior of the coherent spectrum in the ` large ' @xmath229 domain @xmath286 . although ( [ large - x ] ) arises from smaller impact parameters than those contributing to ( [ small - x ] ) , we can still verify the consistency of the assumption @xmath163 used throughout our study . indeed , here we have @xmath287 , implying @xmath288 . although @xmath268 $ ] is not exactly a scaling function of @xmath277 ( in particular when @xmath276 ) , we consider the function [ sar - interpolation ] s_appr[ ] = as a simple approximation to @xmath268 $ ] . in the small @xmath257 limit , ( [ sar - interpolation ] ) has the same parametric behavior as @xmath268 $ ] ( see ( [ s - log - int ] ) ) . at large @xmath257 , both @xmath289 $ ] and @xmath255 $ ] behave as @xmath290 , with however the normalization of @xmath289 $ ] being enhanced ( by a factor @xmath291 ) when compared to @xmath255 $ ] ( see ( [ large - omega ] ) ) . the numerical accuracy of ( [ sar - interpolation ] ) is illustrated in fig . [ fig : appr - scaling ] ( right ) . the deviations of @xmath289 $ ] ( dashed lines ) with respect to @xmath255 $ ] ( solid lines ) are limited to roughly a percent when @xmath292 and to @xmath293 at @xmath294 . when @xmath295 , @xmath289 $ ] has a correct shape but not the correct normalization , and the exact expression ( [ sar ] ) should be preferred . $ ] defined in ( [ sar ] ) with @xmath280 , for various values of @xmath296 . right : @xmath268 $ ] ( solid lines ) compared to @xmath297 $ ] ( dashed lines ) for various values of @xmath296 . for clarity the curves at a given @xmath296 are multiplied by a factor @xmath298.,title="fig:",width=268 ] $ ] defined in ( [ sar ] ) with @xmath280 , for various values of @xmath296 . right : @xmath268 $ ] ( solid lines ) compared to @xmath297 $ ] ( dashed lines ) for various values of @xmath296 . for clarity the curves at a given @xmath296 are multiplied by a factor @xmath298.,title="fig:",width=268 ] in summary , when @xmath299 is large enough , @xmath289 $ ] is a good approximation to @xmath255 $ ] up to values @xmath300 , and the spectrum ( [ spec - alln ] ) can thus be approximated by [ spec - alln - appr ] x x ( 2c_r - n_c ) . let us remark that the logarithm in the expression ( [ spec - alln - appr ] ) at all orders in opacity ( @xmath66 ) can be formally obtained from the logarithm in the spectrum ( [ specn1 ] ) at first order in opacity ( @xmath301 ) by replacing the typical transverse exchange in a single scattering @xmath302 by the typical exchange acquired in multiple _ coulomb _ rescattering , given by @xmath303 when @xmath66 @xcite . the overall factor @xmath304 in ( [ specn1 ] ) can be interpreted as the rescattering probability , @xmath305 , which becomes unity when @xmath66 . ] as already mentioned in section [ sec : intro ] , in refs . @xcite the result ( [ spec - aps-0 ] ) was obtained ( in the case @xmath67 ) from a calculation at first order in opacity , by a similar formal replacement . the above calculation shows that this procedure actually yields a correct approximation to the exact coherent spectrum ( [ spec - alln ] ) at moderate @xmath229 , provided @xmath68 in ( [ spec - aps-0 ] ) is understood as the _ typical _ exchange in multiple coulomb scattering @xmath306 . at large @xmath229 however , the result ( [ spec - aps-0 ] ) ( with @xmath306 ) overestimates by a factor @xmath307 the limiting behavior ( [ large - x ] ) of the exact spectrum ( [ spec - alln ] ) . as already mentioned , deriving the large @xmath229 limit ( [ large - x ] ) requires working _ beyond _ the harmonic oscillator approximation . in particular , the spectrum ( [ spec - armesto2 ] ) ( corresponding to the case of @xmath2 scattering mediated by color singlet @xmath35-channel exchange ) , which can be shown to ensue from ref . @xcite where the harmonic oscillator approximation is used , does not capture either the proper large @xmath229 limit of the spectrum . note that ( [ spec - armesto2 ] ) coincides in ( [ spec - armesto2 ] ) as the typical transverse exchange in the target nucleus a. ] with ( [ u - int ] ) , which as we discussed holds only in the small @xmath229 region ( see the discussion after ( [ u - int ] ) ) . the average coherent energy loss associated to the exact spectrum ( [ spec - alln ] ) reads [ eloss - all - n ] p^+ = p^+ _ 0^ dx x = ( 2 c_r - n_c ) p^+ _ 0^ d s[;r ] . when @xmath66 , the first moment of @xmath255 $ ] is obtained starting from ( [ sar - appr ] ) and using the identities @xmath308 and @xmath309 , [ firstmoment ] _ 0^ d s[;r ] . we thus obtain [ final - eq ] p^+ _ r 1 ( 2 c_r - n_c ) p^+ . similarly to the result ( [ elossn1 ] ) obtained at first order in opacity , the average energy loss is proportional to @xmath24 . the integral in ( [ eloss - all - n ] ) is dominated by @xmath310 , _ i.e. _ , @xmath311 . this stresses that our calculation assuming soft radiation ( @xmath64 ) is consistent provided the accumulated transfer @xmath312 is smaller than @xmath165 . note that the calculation of the average loss using the approximation ( [ sar - interpolation ] ) instead of @xmath268 $ ] would overestimate the exact result ( [ final - eq ] ) by a factor @xmath313 . this is because in the region @xmath310 ( equivalently @xmath314 ) , deviations of @xmath289 $ ] with respect to @xmath268 $ ] are formally of order @xmath315 . up to now the hard process was chosen as the small angle ( but large enough @xmath165 ) scattering of a fast parton , either quark or gluon , see fig . [ fig : gbamplitude]a . in this case the fully coherent induced radiation derived above can be interpreted as the radiative energy loss of an asymptotic quark or gluon . the coherent radiation originates dominantly from the interference between initial and final state radiation ( see fig . [ fig : abcn1]c ) , and should arise independently of the hard process as long as the energetic incoming and outgoing particles are colored . to illustrate this , we study in this section two examples where the nature of the charged particle is modified by the hard process . first , we consider the case of an incoming fast quark being scattered at small angle ( in the target rest frame ) to an outgoing fast gluon through color - triplet @xmath35-channel exchange , see fig . [ fig : other - hard]a . second , we study the scattering of a fast gluon to a fast pointlike color octet state of mass @xmath37 via single gluon exchange , see fig . [ fig : other - hard]b . this case is suited to the production of an octet heavy quark pair in gluon - gluon fusion , @xmath316_{\rm 8}$ ] , in the approximation where the @xmath317 pair is compact . in those examples the coherent radiation arises from the interference between emission amplitudes off different objects , and can not be interpreted as the energy loss of a well - defined particle . it might be called ` parton energy loss ' by abuse of language , but it is simply the medium - induced radiation associated to a given hard process . -channel exchange . ( b ) scattering of an energetic gluon to an energetic massive pointlike color octet state ( denoted by the dashed line ) via single gluon exchange.,width=377 ] focussing as before on radiation with large formation time @xmath27 , the medium - induced radiation spectrum associated to the hard process of fig . [ fig : other - hard]a is given , at first order in opacity , by some sets @xmath190 , @xmath191 , @xmath192 of diagrams analogous to those of fig . [ fig : abcn1 ] . when @xmath27 , the sets @xmath190 and @xmath191 corresponding to purely final and initial state radiation vanish , and what remains is the set @xmath192 of interference diagrams . in the present situation we must distinguish the cases where the rescattering center located at @xmath194 is before or after the hard scattering vertex . for @xmath318 , what rescatters is a color octet ( see fig . [ fig : other - hard]a ) , giving a contribution to the spectrum similar to ( [ spec - order-1 ] ) , . z_h = d^2 k_z_h^l d^2 v ( ) , [ z1>zh ] where the set @xmath319 is given in fig . [ fig : cacb]a and @xmath320 . for @xmath321 , what rescatters is a color triplet , yielding the contribution . < z_h = d^2 k_0^z_h d^2 v ( ) , [ z1<zh ] where the set @xmath322 is given in fig . [ fig : cacb]b and @xmath323 . note the proper normalization with respect to the quark mean free path @xmath324 in ( [ z1<zh ] ) . applying the pictorial rules of fig . [ fig : pictorial - rules ] we easily obtain = = . using @xmath325 , the contributions from @xmath318 and @xmath321 add up to x = n_c d^2 kd^2 v ( ) . [ spec - qtog ] the medium - induced spectrum associated to the hard process of fig . [ fig : other - hard]a is thus identical to the spectrum ( [ spectrum - n1 ] ) associated to the hard scattering of an asymptotic gluon.the fact that those spectra have an identical color factor @xmath130 is simply understood as follows . in general , the overall color factor of the fully coherent radiation , due to the structure of the interference term , is of the form @xmath326 , where the incoming and outgoing particles are in the color representations @xmath327 and @xmath328 respectively . using @xmath329 , where @xmath330 is the color charge of the @xmath35-channel exchange , we recover the factor @xmath331 in the case of asymptotic gluon scattering considered in section [ sec : hard - smallangle ] , and the factor @xmath332 in the present case . ] the spectrum ( [ spec - qtog ] ) and associated radiative loss were evaluated analytically in section [ sec : n1qcd ] , see ( [ specn1 ] ) and ( [ elossn1 ] ) . as expected , a fully coherent radiative loss proportional to @xmath24 arises despite different ( but non - zero ) initial and final color charges in the process of fig . [ fig : other - hard]a . the above results trivially extend to all orders in opacity , as well as to the process of a fast gluon scattered to a fast quark . and @xmath322 of diagrams contributing to the radiation spectrum when @xmath318 and @xmath321 respectively ( see ( [ z1>zh ] ) and ( [ z1<zh ] ) ) , as well as color factors @xmath333 and @xmath334.,width=574 ] we now consider a hard process where an octet state of mass @xmath37 is produced in gluon - gluon fusion , see fig . [ fig : other - hard]b . for soft gluon radiation with @xmath27 , the precise partonic content of the octet state ( @xmath317 , @xmath335 , ... ) as well as the internal structure of the hard process producing it ( the blob in fig . [ fig : other - hard]b ) is irrelevant as long as the octet state is compact enough . within this assumption the amplitude for soft gluon emission off the final state is effectively the same as off a pointlike , massive gluon ( the dashed line in fig . [ fig : other - hard]b ) . the calculation of the fully coherent radiation spectrum is thus identical to the case of an asymptotic gluon , up to the replacement [ insert - mass ] to account for the mass dependence of the gluon emission vertex off the final state . at first order in the opacity expansion , we start from the expression ( [ spectrum - n1 ] ) modified accordingly , x = n_c d^2 kd^2 v ( ) . [ spec - massive - octet ] the explicit calculation is performed along the same lines as in section [ sec : n1qcd ] . we first average over the azimutal angles of @xmath208 and @xmath61 , = ( ^2 - k^2 ) , = ( k^2 ) , where @xmath336 , to obtain x = n_c _ 0^ d^2 v ( ) _ 0^^2 ( k^2 ) . [ spec - massive - octet-2 ] integrating by parts yields x = n_c _ 0^ d^2 ( ^2 ) . [ spec - massive - octet-3 ] with the change of variable @xmath337 we arrive at x = n_c ( _ , ) ; _ ; , [ spec - massive - octet - n1 ] where the function @xmath338 is defined as ( _ , ) _ 0^ . the integral over @xmath266 can be evaluated analytically , providing an exact expression of the spectrum ( [ spec - massive - octet - n1 ] ) . however , a simple approximation to the spectrum can be obtained by examining the limiting behaviors of @xmath339 . for @xmath340 and fixed @xmath341 ( note that @xmath342 ) , the integral over @xmath266 defining @xmath339 is dominated by the logarithmic interval @xmath343 , leading to [ sigma - limit1 ] ( _ , ) _ _ 1 . when @xmath344 , the @xmath266-integral can be shown to be dominated by @xmath345 , [ sigma - limit2 ] ( _ , ) _ _ 1 _ 1/_^2 ^ 1 . those limiting behaviors suggest the simple approximation ( _ , ) , which reproduces the leading parametric behaviors of @xmath339 in the limits @xmath340 and @xmath344 . the spectrum ( [ spec - massive - octet - n1 ] ) can thus be approximated as [ spec - massive - octet - n1-appr ] x n_c . this result is simply obtained from the spectrum ( [ specn1 ] ) by replacing the hard scale @xmath165 ( in the @xmath346 case of section [ sec : n1qcd ] ) by @xmath347 . the spectrum ( [ spec - massive - octet - n1-appr ] ) is not exact ( except when @xmath346 ) , but as mentioned above exhibits a correct parametric dependence . in particular it yields the average radiative loss p^+ p^+ dx x = n_c _ s p^+ , [ eloss - massive - appr ] to be compared to the exact loss derived from ( [ spec - massive - octet - n1 ] ) , p^+ = n_c _ s p^+ ( ) ; ( ) _ 0^ ( y , ) . [ eloss - massive - octet - n1-exact ] eqs . ( [ eloss - massive - appr ] ) and ( [ eloss - massive - octet - n1-exact ] ) differ only by the factor @xmath348 , which can be checked to be a monotonous function of @xmath349 , with values @xmath350 and @xmath351 at @xmath352 and @xmath353 respectively . thus , the factor @xmath348 does not affect the parametric dependence of ( [ eloss - massive - appr ] ) . at first order in opacity , the medium - induced radiation associated to the production of a heavy color octet state ( fig . [ fig : other - hard]b ) is parametrically similar to that associated to the scattering of an energetic ( massless ) gluon studied in section [ sec : n1qcd ] , up to the change in the hard scale , @xmath354 . below we verify that this statement still holds when resumming all orders in opacity . at all orders in opacity , the spectrum is given by the expression ( [ spec - alln - rescaled ] ) modified according to ( [ insert - mass ] ) , x = n_c d^2 k . [ spec - alln - mass ] the spectrum can be derived as in section [ sec : exact - der ] . only one additional identity is needed , d^2 k e^- i kb = _ 0(b ) xmb k_1(xmb ) . the final result is x = n_c s [ ; r ; _ _ m ] ; ; r ; _ _ m , 1.5 cm & & [ spec - alln - mass-2 ] + s [ ; r ; _ _ m ] = 2 _ 0^ j_0(b ) \ { 1 - } _ _ mb k_1(__m b ) . & & [ somegam ] comparing ( [ spec - alln - mass-2 ] ) and ( [ somegam ] ) to ( [ spec - alln ] ) and ( [ sar ] ) , we see that the presence of the mass parameter @xmath37 introduces the factor @xmath355 , which as expected tends to unity when @xmath48 . using the same procedure as in section [ useful - appr ] , we find that at fixed @xmath66 , the function @xmath356 $ ] can be approximated by [ massive - s - appr ] s [ ; r ; _ _ m ] s_appr [ ; r ; _ _ m ] , where @xmath357 was introduced before . thus the spectrum ( [ spec - alln - mass-2 ] ) takes the simpler form [ massive - spec - appr ] x n_c . as was the case at first order in opacity ( see previous section ) , at all orders in opacity the spectrum ( [ massive - spec - appr ] ) is obtained from that of an energetic massless gluon ( [ spec - alln - appr ] ) by replacing @xmath354 . the result ( [ massive - spec - appr ] ) corresponds to the spectrum derived semi - heuristically in @xcite and @xcite , and used for phenomenology in @xcite . the numerical accuracy of ( [ massive - s - appr ] ) is shown in fig . [ fig : massive - s - appr ] for various values of @xmath296 and of the variable @xmath358 . the case @xmath352 ( @xmath346 ) was studied in section [ useful - appr ] , see fig . [ fig : appr - scaling ] . $ ] defined in ( [ somegam ] ) ( solid lines ) compared to the approximation @xmath359 defined in ( [ massive - s - appr ] ) ( dashed lines ) , as a function of the variable @xmath360 , for different values of @xmath296 ( @xmath361 ) and of @xmath358 : @xmath362 ( left ) , @xmath363 ( middle ) , and @xmath364 ( right).,title="fig:",width=185 ] $ ] defined in ( [ somegam ] ) ( solid lines ) compared to the approximation @xmath359 defined in ( [ massive - s - appr ] ) ( dashed lines ) , as a function of the variable @xmath360 , for different values of @xmath296 ( @xmath361 ) and of @xmath358 : @xmath362 ( left ) , @xmath363 ( middle ) , and @xmath364 ( right).,title="fig:",width=185 ] $ ] defined in ( [ somegam ] ) ( solid lines ) compared to the approximation @xmath359 defined in ( [ massive - s - appr ] ) ( dashed lines ) , as a function of the variable @xmath360 , for different values of @xmath296 ( @xmath361 ) and of @xmath358 : @xmath362 ( left ) , @xmath363 ( middle ) , and @xmath364 ( right).,title="fig:",width=185 ] in hard processes where either the incoming or the outgoing energetic particle is colorless ( see fig . [ fig : no - interference ] ) , the interference responsible for the fully coherent radiation is absent , and the induced radiation reduces to purely final or purely initial state radiation . in the present section we briefly review the latter contributions and recover known results . for simplicity we work at first order in the opacity expansion . the calculations at all orders in opacity @xcite do not bring any important qualitative change . as we have seen in section [ sec : n1qcd ] , initial and final state radiation cancels out in the ( @xmath59-integrated ) radiation spectrum in the limit @xmath27 , see ( [ setan1 ] ) and ( [ setbn1 ] ) . this means that assuming @xmath27 is too drastic to derive initial / final state radiation . the purely initial / final average energy loss turns out to be dominated by @xmath365 , scaling as @xmath366 and independent of @xmath24 ( up to logarithms ) , see ( [ fs - eloss ] ) . this type of energy loss , @xmath366 , received much attention in the last two decades . we emphasize that in situations where the interference between initial and final state radiation can not be neglected ( as studied in the previous sections ) , @xmath366 is actually negligible at large enough @xmath24 , compared to the fully coherent loss @xmath23 . hard process . ( b ) fast quark annihilation in the @xmath367 hard process.,width=377 ] a toy model for a process where only final state radiation contributes is shown in fig . [ fig : no - interference]a . here the hard transfer @xmath61 is irrelevant and we can choose a frame where the final energetic quark has zero transverse momentum . at order @xmath189 in the opacity expansion , the spectrum for final state radiation off an energetic quark is given by ( compare to ( [ spec - order - n ] ) ) . = d^2 k d^2 v ( ) , [ fs - spec - order-1 ] where only one generic contribution is shown is the sum of diagrams . since we do not assume @xmath27 any longer , diagrams where the gluon emission time @xmath35 in the amplitude ( or @xmath179 in the conjugate amplitude ) occurs within the target , @xmath368 , can not be neglected . such diagrams are proportional to a difference of phase factors @xmath369 which can be derived in time - ordered perturbation theory @xcite . taking into account those phase factors , the rules of fig . [ fig : pictorial - rules]a for gluon emission vertices must be modified to the rules defined in fig . [ fig : pictorial - rules - phases ] . ) and ( b ) initial state ( @xmath321 ) emission vertices supplemented by phase factors . for convenience we set @xmath370 , and we define @xmath371 , @xmath372.,width=491 ] as can be easily verified , among the diagrams contributing to ( [ fs - spec - order-1 ] ) , those where @xmath35 and @xmath179 belong to the same time - interval ( _ i.e. _ , either @xmath373 or @xmath374 ) give a vanishing contribution to the @xmath59-integrated radiation spectrum . this property was already mentioned in section [ sec : n1qcd ] . the remaining diagrams sum up toa ) is not drawn in eq . ( [ fig : fs - diagrams ] ) . ] & & -3 mm = + & & 2 n_c c_f \ { ( 1- ) . + . ( 1- ) } + & & = 2 n_c^2 c_f \ { ( 1- ) - ( 1- ) } . [ fig : fs - diagrams ] inserting the latter result in ( [ fs - spec - order-1 ] ) and shifting variable @xmath375 in the first term we get . \|_fs = d^2 k_z_h^l d^2 v ( ) 2 n_c ( 1- ) . finally , integrating over @xmath194 and using @xmath325 yields . x \|_fs = d^2 kd^2 v ( ) , [ fs - spec - order-1-vitev ] where @xmath6 is the distance crossed by the quark from the hard production point @xmath370 to the boundary of the medium . clearly , the spectrum for purely final state radiation off a color charge @xmath50 is obtained by replacing @xmath376 in the expression ( [ fs - spec - order-1-vitev ] ) , which then coincides with the final state radiation spectrum derived in ref . @xcite . the @xmath59-integrated spectrum ( [ fs - spec - order-1-vitev ] ) is mathematically well - defined , in particular there is no singularity at @xmath112 or @xmath377 . changing variable @xmath375 , ( [ fs - spec - order-1-vitev ] ) becomes . = d^2 v ( ) . [ fs - spec - order-1-vitev-2 ] the integral over @xmath208 can be performed using = = 2 ( ^2 - k^2 ) , and introducing the variable @xmath378 we obtain . x \|_fs = _ 0^ du . when @xmath379 , the @xmath266-integral simplifies to @xmath380 , giving . x \|_fs = ( x p^+ ^2 l ) . the spectrum arises from an integration domain where @xmath381 , _ i.e. _ , from formation times @xmath382 , as expected . to logarithmic accuracy the associated average energy loss reads . p^+ interval @xmath383 . the result ( [ fs - eloss ] ) was derived previously , see for instance ref . when @xmath213 the purely final state energy loss is independent of @xmath24 ( up to logarithms ) and is proportional to @xmath384 . in the case where the outgoing energetic particle participating to the hard process carries no color charge , as in fig . [ fig : no - interference]b , the medium - induced radiation arises solely from the initial state . the derivation of the spectrum is the same as for final state radiation , up to few modifications . the diagrams contributing to the initial state radiation spectrum . x \|_is = d^2 k d^2 v ( ) [ is - spec - order-1 ] are evaluated using the pictorial rules defined in fig . [ fig : pictorial - rules - phases]b , which differ from those of fig . [ fig : pictorial - rules - phases]a only in the phase factors . the only non - vanishing contribution to the @xmath59-integrated spectrum ( [ is - spec - order-1 ] ) arises from the set of diagrams & & -3 mm = + & & 2 n_c c_f \ { ( 1- ) . + . ( 1- ) } + & & = 2 n_c^2 c_f ( 1- ) . [ fig : is - diagrams ] as a result , . = d^2 kd^2 v ( ) , [ is - spec - order-1-end ] where @xmath6 is the distance travelled by the quark in the medium up to the point @xmath370 . thus , provided the radiated gluon transverse momentum @xmath59 is integrated over , the purely initial and purely final state radiation spectra are identical , see ( [ fs - spec - order-1-vitev-2 ] ) and ( [ is - spec - order-1-end ] ) . the associated average energy loss is given by ( [ fs - eloss ] ) . in this study we have derived , to all orders in the opacity expansion , and for the case of a coulomb rescattering potential , the _ medium - induced _ gluon radiation spectrum associated to the hard forward scattering of an energetic parton . the latter spectrum arises from gluon formation times scaling as the parton energy , @xmath385 , corresponding to _ fully coherent _ radiation ( _ i.e. _ @xmath27 ) over the medium size . in the light - cone gauge @xmath90 , the induced coherent spectrum is dominated by the interference between initial and final state emission amplitudes . using the result ( [ spec - alln - appr ] ) for a parton of color charge @xmath50 ( process @xmath2 or @xmath1 ) , the demonstration in section [ sec : qtog ] that ( [ spec - alln - appr ] ) also holds ( up to the color factor ) for processes where the nature of the fast parton is changed in the hard scattering ( @xmath3 and @xmath4 ) , and finally the result ( [ massive - spec - appr ] ) for the scattering of a gluon to a _ massive _ compact color octet , we infer the following general expression for the induced coherent spectrum , [ spec - general ] x = ( c_r + c_r - c_t ) . the color factor trivially follows from the structure of the interference term . the latter is indeed of the form @xmath326 ( where @xmath119 and @xmath386 are the color generators of the incoming and outgoing parton color representations @xmath327 and @xmath328 ) , which can be written as [ color - rule ] 2 t_r^a t_r^a = ( t_r^a)^2 + ( t_r^a)^2 - ( t_r^a - t_r^a)^2 = c_r + c_r - c_t , with @xmath50 , @xmath51 the incoming and outgoing color charges , and @xmath52 the color charge of the @xmath35-channel exchange . we emphasize that the factor @xmath387 holds at any finite @xmath130 , and is negative in the particular @xmath2 case , where @xmath388 . we argued in section [ sec : qdep ] that a medium - induced energy _ gain _ for the @xmath2 process , although surprising , could have been anticipated from the features of the _ total _ radiation spectrum associated to @xmath2 . note however that the negativity of the induced spectrum associated to @xmath2 is in general not a specificity of the quark , but depends on the @xmath35-channel color exchange @xmath52 . in particular , the induced spectrum associated to @xmath1 would also be negative in an academic situation where @xmath52 would be constrained to satisfy @xmath389 . a natural question to ask is whether this negative color factor could have important phenomenological consequences . although phenomenology is not the purpose of this paper , let us mention a couple of observables which could be sensitive to an _ energy gain _ arising from a negative medium - induced gluon spectrum . the production of photons in proton nucleus ( p a ) collisions is likely dominated by the compton scattering channel , @xmath390 , at forward rapidity . although this process is formally @xmath391 , it was shown in @xcite that @xmath392 should be taken as the _ global _ color charge of the 2-particle final state ; consequently , the color factor of the medium - induced gluon spectrum in @xmath393 and @xmath390 should be identical and equal to @xmath131 . similarly , forward light - hadron production should be sensitive to the scattering channel @xmath394 , and therefore be sensitive to negative medium - induced gluon spectrum when the final @xmath395-state is in the color triplet representation ( unlike @xmath396 , the @xmath395-state can be in other , higher - dimension , color representations ) . in both channels , @xmath397 and @xmath398 , one could legitimately expect some enhancement in p a with respect to p p collisions , although this remains to be quantified . in particular , the smallness of the color factor @xmath131 may somehow tame such an enhancement . the induced coherent spectrum explicitly depends on the hard process ( through @xmath52 and the hard scale @xmath399 ) . when the incoming particle is colorless ( @xmath400 ) , the interference between initial and final state emission amplitudes is absent and fully coherent radiation vanishes ( which formally follows by noting that when @xmath400 , color conservation implies @xmath401 ) . in this case , we verified in section [ sec : is - fs ] ( at first order in the opacity expansion ) that the spectrum found in our setup arises from formation times @xmath402 and coincides with previously known results @xcite . strictly speaking , ( [ spec - general ] ) is an approximation to the exact spectrum , which however has the correct parametric behavior when @xmath403 ( see sections [ useful - appr ] and [ sec : massive - all - orders ] ) . in the case of a coulomb scattering potential considered in our study , the _ typical _ transverse broadening entering ( [ spec - general ] ) reads @xmath404 at large @xmath6 ( specifically when @xmath405 , see @xcite ) . we however found that ( [ spec - general ] ) is _ numerically _ a very accurate approximation to the exact coherent spectrum as soon as @xmath406 , _ i.e. _ already at moderate values of @xmath6 , and up to values of @xmath407 . we have seen that at larger @xmath229 , @xmath408 , the parametric behavior @xmath409 of ( [ spec - general ] ) does not reproduce the proper normalization of the exact spectrum . in the massless case , it overestimates ( by a factor @xmath307 ) the correct behavior @xmath410 , see ( [ large - x ] ) and fig . [ fig : appr - scaling ] ( right ) . in the massive case , the approximation ( [ spec - general ] ) is also inaccurate at large @xmath229 , see fig . [ fig : massive - s - appr ] . in the large @xmath229 domain the exact expression of the spectrum ( eqs . ( [ spec - alln - mass-2])([somegam ] ) in the massive case ) should be preferred . we stress that the spectrum ( [ spec - general ] ) corresponds to the induced coherent radiation in a target of finite size @xmath6 with respect to an ideal target of zero size . for the sake of the following discussion , we introduce the induced spectrum in p a collisions with respect to p p collisions , defined as the difference of the spectrum ( [ spec - general ] ) for a target nucleus a and a proton target , [ spec - general - pavspp ] . x \|_pa - pp = ( c_r + c_r - c_t ) . in some aspects , our study of coherent gluon radiation resembles some studies of gluon radiation in the saturation framework , however with a crucial difference : in the saturation formalism , the focus is usually on that part of gluon production which is _ process - independent _ ( and related to the _ unintegrated _ or _ @xmath411-dependent _ gluon distribution in the target ) , whereas we concentrated here on the _ process - dependent _ part of gluon radiation . to illustrate this point , it is instructive to compare our study to the study of gluon production in nuclear dis and p a collisions of ref . @xcite . in this work , the focus is on the contributions to the _ total _ , @xmath132-differential radiation spectrum which exhibit a _ collinear _ ( logarithmic ) singularity when @xmath412 . as a consequence , gluon production in p a collisions is dominated ( in light - cone gauge ) by purely initial and purely final state radiation . the interference term , ) corresponds to equation ( 60 ) of ref . @xcite and represents the contribution of the interference to the total gluon radiation spectrum off a fast massless quark in p a collisions . ] [ mk60 ] . x \|_pa , int = -2 c_f , is non - logarithmic when @xmath412 and is thus neglected in @xcite . however , considering _ medium - induced _ instead of _ total _ radiation in the setup of ref . @xcite , it is straightforward to check that purely initial and final state radiation ( integrated over @xmath45 ) cancels out ( as in the present study ) , leaving only a contribution from the interference term . thus , in ref . @xcite the medium - induced spectrum can be simply obtained by subtracting from ( [ mk60 ] ) a similar contribution in p p collisions , [ mk60-induced ] . x \|_pa - pp = -2 c_f . integrating ( [ mk60-induced ] ) over @xmath45 yields [ mk60-induced-2 ] . x \|_pa - pp = 2 c_f , arising from the domain @xmath413 . this emphasizes that the induced contribution ( [ mk60-induced-2 ] ) lies beyond the scope of the @xmath412 limit ( which formally implies @xmath414 ) used in ref . @xcite . quite remarkably , ( [ mk60-induced-2 ] ) can be obtained from the general expression ( [ spec - general - pavspp ] ) by the formal replacements @xmath415 , @xmath416 ( and @xmath417 , since ( [ mk60-induced-2 ] ) corresponds to the scattering of a fast quark ) . ) can also be obtained ( including the color factor ) _ directly _ from the @xmath418 limit of the expression ( [ spec - armesto2 ] ) ( which expression follows from the setup of ref . @xcite ) . indeed , as mentioned in @xcite , when @xmath418 the setups of @xcite and @xcite coincide . this is because both studies consider the @xmath2 process , and @xmath419 in the model of ref . ] this can be understood from the fact that there is no ` hard process ' in the setup of ref . @xcite , the fast quark undergoing only soft rescatterings in the nuclear target . we are most grateful to al mueller and stphane munier for useful exchanges , and to tseh liou and al mueller for a very rich and instructive correspondence on the similarities between their recent work @xcite and our study . despite different theoretical formalisms and setups used in @xcite and here , our studies agree on important points , in particular on the parametric dependence of the medium - induced soft gluon radiation spectrum . feynman diagrams have been drawn with the jaxodraw software @xcite . this work is funded by `` agence nationale de la recherche '' under grant anr - partonprop . in this appendix we briefly outline the derivation of eq . . the soft radiation _ intensity _ accompanying the ( hard ) production of a parton of momentum @xmath420 is given by the ratio di = , [ app : di - general ] where the numerator is the radiative cross section for producing a two - parton final state ( the leading parton @xmath420 and a soft radiated gluon of momentum @xmath421 ) while the denominator represents the inclusive cross section for producing the leading parton . formally the latter inclusive cross section includes the two - parton production cross section entering the numerator , integrated over the radiated gluon phase space . however , to leading order in @xmath422 , the denominator can be approximated by the production cross section of the leading parton _ without _ soft gluon emission . moreover , assuming that soft rescatterings of the leading parton do not affect @xmath423 , the latter can be calculated as if there were no scattering center in the medium , _ i.e. _ , at zeroth order in the opacity expansion @xcite . on the contrary , the numerator of is modified by soft rescatterings and must be represented as an expansion in the number @xmath166 of scattering centers encountered by the leading parton . the @xmath424 term in this expansion coincides with the radiation cross section in the absence of a medium , and thus cancels out in the _ medium - induced _ radiation intensity defined as di^induced di - di^vacuum = _ n=1^ _ n=1^ di^(n ) . [ app : di - induced ] each term in the sum over @xmath166 contains an implicit average over the positions of the scattering centers on which the parton - gluon system rescatters . assume there are @xmath425 scattering centers overall , which are randomly but , on the average , uniformly distributed in a layer of thickness @xmath6 and transverse area @xmath426 . ( recall that @xmath173 is the screening length of the coulomb scattering potential ( [ coulomb - pot ] ) . ) in the calculation we consider the interaction of each scattering center to the lowest order in @xmath422 . this implies that each of the @xmath166 scattering centers either contributes a single born interaction in both the amplitude and the conjugate , or it can contribute a double born interaction in the amplitude with no interaction in the conjugate amplitude ( or vice versa ) , thus providing a so - called virtual correction to the aforementioned single - born interaction . averaging over the position of each scattering center introduces a factor @xmath427 . accounting the different possibilities to choose @xmath428 scattering centers contributing via single born interaction together with @xmath429 and @xmath430 centers contributing double born interactions in the amplitude and conjugate amplitude respectively , brings a factor @xmath431 . integrating over @xmath432 ensures the same transverse momentum transfer from the scattering center @xmath433 in the amplitude and its conjugate for a single born scattering , or a zero net momentum transfer in a double born interaction @xcite . for @xmath434 one can use @xmath435 to obtain di^(n ) = _ n_1 , n_2 , n_\|2 , [ app : dsigma - n - averaged ] where @xmath436 is the number of scattering centers per unit volume in the target , @xmath437 , and @xmath438 . the next observation concerns the square of the matrix element entering the cross section @xmath439 in . in our setup of small angle scattering and gluon radiation with large formation times , each rescattering contributes to the lorentz structure a factor equal to the elastic scattering cross section @xmath440 of an ` abelian ' parton with charge @xmath81 . this factor is parametrized as = _ el^abelian v(_i ) , where @xmath172 is the normalized coulomb potential defined in ( [ coulomb - pot ] ) . the elastic cross section which corresponds to the hard scattering cancels the denominator except for the color factor @xmath441 . therefore di^(n ) = _ n_1 , n_2 , n_\|2 ( 2 g)^2 , where the phase space factor for the radiated gluon ( see ( [ phase - space - factor ] ) ) and a factor @xmath442 ( see ( [ gb ] ) ) have been made explicit , and the ` reduced cross section ' @xmath443 contains only the color factors and emission vertices computed using the pictorial rules defined in fig . [ fig : pictorial - rules ] . introducing light - cone time ( or longitudinal position ) ordering of the scattering centers , @xmath444 for @xmath445 , the permutations of the scattering centers within the subsets of those contributing via single born or double born interaction ( in either the amplitude or its conjugate ) give @xmath446 identical contributions . noting that @xmath447 completes the proof of ( [ spec - order - n ] ) . here we derive the behavior of the function @xmath255 $ ] defined in ( [ sar ] ) in the limit @xmath285 at fixed @xmath296 . the function @xmath255 $ ] can be written as [ sar - alpha-1 ] s[;r ] = _ 0^ db b j_0(b ) ( b ) , where [ alphab ] ( b ) \ { 1 - } . using b j_0(b ) = b j_1(b ) a simple integration by parts of ( [ sar - alpha-1 ] ) yields [ sar - alpha-2 ] s[;r ] = - _ 0^ db j_1(b ) b , where @xmath448 . now use _ 1(b ) = ( 1- j_0(b ) ) to integrate ( [ sar - alpha-2 ] ) by parts , leading to [ sar - alpha-3 ] s[;r ] & = & _ 0^ db ( b) + & = & - _ 0^ db j_0(b ) ( b) . we used @xmath449 , which can be verified from ( [ alphab ] ) . when @xmath285 , the integral in the second term of ( [ sar - alpha-3 ] ) vanishes due to the rapid oscillation of @xmath278 . we thus obtain s[;r ] _ . [ app : large - omega ] 31ifxundefined [ 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/physrevlett.91.172302 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2013.01.024 [ * * , ( ) ] , link:\doibase 10.1016/j.physletb.2010.12.020 [ * * , ( ) ] , link:\doibase 10.1140/epjc / s10052 - 012 - 1945-x [ * * , ( ) ] , + link:\doibase 10.1016/s0550 - 3213(96)00553 - 6 [ * * , ( ) ] , link:\doibase 10.3367/ufne.0179.200907a.0697 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.83.114036 [ * * , ( ) ] , @noop * * , ( ) , @noop * * , ( ) , link:\doibase 10.1007/jhep05(2013)155 [ * * , ( ) ] , + + link:\doibase 10.1016/j.physletb.2012.09.039 [ * * , ( ) ] , link:\doibase 10.1007/jhep12(2013)052 [ * * , ( ) ] , link:\doibase 10.1016/s0550 - 3213(98)00384 - 8 [ * * , ( ) ] , + link:\doibase 10.1103/physrevd.25.746 [ * * , ( ) ] in @noop _ _ ( ) pp . , link:\doibase 10.1103/physrevd.22.2157 [ * * , ( ) ] in @noop _ _ ( ) in @noop _ _ ( ) @noop * * , ( ) , link:\doibase 10.1016/0550 - 3213(94)90079 - 5 [ * * , ( ) ] , + @noop * * , ( ) , + link:\doibase 10.1016/0550 - 3213(74)90154 - 0 [ * * , ( ) ] link:\doibase 10.1016/0370 - 2693(93)91724 - 2 [ * * , ( ) ] , + link:\doibase 10.1016/s0550 - 3213(96)00581 - 0 [ * * , ( ) ] , link:\doibase 10.1134/1.1358417 [ * * , ( ) ] , link:\doibase 10.1103/physrevc.75.064906 [ * * , ( ) ] , @noop * * , ( ) , link:\doibase 10.1007/jhep01(2015)141 [ * * , ( ) ] , link:\doibase 10.1103/physrevd.89.074026 [ * * , ( ) ] , + link:\doibase 10.1016/j.cpc.2009.02.020 [ * * , ( ) ] ,	we revisit the medium - induced coherent radiation associated to _ hard _ and _ forward _ ( small angle ) scattering of an energetic parton through a nuclear medium . we consider all @xmath0 hard forward processes ( @xmath1 , @xmath2 , @xmath3 and @xmath4 ) , and derive the energy spectrum of induced coherent radiation rigorously to all orders in the opacity expansion and for the specific case of a coulomb scattering potential . we obtain a simple general formula for the induced coherent spectrum , which encompasses the results corresponding to previously known special cases .
quantum machine learning aims at merging the methods from quantum information processing and computer engineering to provide new solutions for problems in the areas of pattern recognition and image understanding @xcite . from one side the research in this area is focused at applying the methods of quantum information processing for solving problems related to classification and clustering in signal processing . from the other perspective the methods for classification developed in computer engineering are used to find solutions for problems like quantum state discrimination @xcite , which are tightly connected with the recent developments in quantum cryptography . the aim of the paper is to present a complete work - flow , based on the mathematical formalism used in quantum information theory , suitable for processing data for the purpose of pattern recognition . the main goal is to introduce the method for describing classical data using entangled quantum states and to investigate the difference between the entangled and the separable representations of states . the presented work - flow enable the processing of real - world data as it can be implemented and executed on standard desktop computers . the main reason why this is possible is the introduction of the quantization procedure which allows the reduction of the dimensionality of the utilized quantum states . this paper is organized as follows . in section [ sec : preliminaries ] we provide some notation and initial considerations related to the problem of using quantum states for the purpose of pattern representation and classification . in section [ sec : quantum - representation ] we propose a general framework for encoding data using the formalism of quantum states and describe the basic classification of the 1d data using this representation . in section [ sec : higher - dims ] we introduce a formalism for constructing representations of feature vectors and provide some examples of using entangled and separable flavors of the introduced formalism in 2d case . finally , in section [ sec : final ] we provide a summary of the presented work and provide concluding remarks . we start by introducing the notation from the area of quantum information relevant for the purpose of processing patterns . we also review the problem of pattern classification and introduce the notation used in the paper . in this paper we use the notation of quantum mechanics in which vectors are written using ket symbols , such that @xmath0 is a base vector with label @xmath1 . in the simplest case of two - dimensional space , which can be used to represent a bit on a quantum computer , we have @xmath2 accordingly , for @xmath3-dimensional space we have @xmath4 with @xmath5 . we also assume that ket vectors are normalized . this notation can be used to express any combination of vectors as @xmath6 usually we require that @xmath7 . let us now assume that we deal with @xmath8 features , which are measured in the experiment . with each feature we connect a vector space @xmath9 of dimension @xmath10 , necessary to store the possible measurement results . a pattern consisting of @xmath8 features can be represented as a tensor product of feature spaces as @xmath11 where @xmath12 denotes tensor ( a.k.a . kronecker ) product @xcite . for two vectors @xmath13 and @xmath14 @xmath15 tensor product is defined as @xmath16 and it is customary to write @xmath17 one of the features of the above formalism is the seamless notation of the scalar product as @xmath18 elements @xmath19 can be written as @xmath20 and they represent scalar products of the base vectors and @xmath21 represents the vector dual to @xmath22 . let us now assume that we are dealing with the problem of the supervised learning . in this situation we have at our disposal a number of representatives associated with labels denoting classes . the classes used to discriminate are constructed using a number of measurement results , each one representing a pattern of a known class @xcite , @xmath23 where number of representatives @xmath24 can be different for each class . the set of class elements with the associated labels is refereed to as the learning set . the classification of the unknown patter @xmath25 is performed using the score function @xmath26 , calculated for the pattern and each class . in the simplest scenario , the score function is calculated as a distance between the unknown pattern and the class representatives . the class representatives @xmath27 for each class @xmath28 are built by the appropriate averaging over the class elements . the unknown pattern is classified according to the greatest value of the score @xmath29 , calculated as the distance between the pattern @xmath25 and the class representatives @xmath27 , @xmath28 , @xmath30 in this section we develop a quantum representation for patterns and classes . we focus on the patterns distributed with the normal distribution , but the provided framework does not depend on this assumption . we start with the simplest case of 1d data and subsequently we consider the possible generalizations of the introduced formalism for the purpose of discrimination patterns described by vectors of features . in order to define a quantum classification procedure one needs to specify the mapping between the space of values and the space of quantum states . this is exactly the counterpart of the quantization function used in machine learning algorithms . however , in this case , this function is used not only to reduce the dimensionality of the problem , but also to obtain a quantum representation of patterns and classes . here we aim at representing data in the form of vectors pure quantum states or corresponding projection operators . in principle one can extend the presented formalism to include mixed quantum states , but the particular choice of pure states enables the straightforward utilization of quantum superposition and quantum entanglement . from the quantum information point of view , the most natural formulation of the pattern classification problem is based on the distance minimization . in distance - based methods we are interested in the distance between the class representative and the pattern we wish to classify @xcite . in order to use quantum states for the purpose of pattern classification we use a framework based on the following assumptions : * patterns are represented as normalized vectors . * classes are represented by the normalized superposition of the representatives . * similarity between the patterns is calculated as a scalar product . the first assumption from the above list requires mapping the classical data on the quantum pure states , @xmath31 in the following the procedure executing this step will be called the quantum quantization and its goal is twofold . firstly , it is required to reduce the feature space and generate the quantization symbols . secondly , it calculates the dimensions of the state space used and maps the quantization symbols onto this space . the quantum quantization step is crucial from the point of view of analyzing the quantum correlations as it effectively defines the structure of the used state space . for a given quantum quantization procedure and a selected distance measure one obtains a complete procedure for quantum - based classification . in order to classify an unknown pattern @xmath25 one has to execute the following steps . * step :* perform quantum quantization by mapping class elements @xmath32 prepare the class representatives @xmath27 for each class @xmath28 by the appropriate averaging over the quantum class elements . calculate the distance between the pattern @xmath25 and the class representatives @xmath27 , @xmath28 , @xmath30 classify the pattern according to the greatest value of @xmath29 . one should note that the class representatives @xmath27 , @xmath28 can be obtained as a flat superposition of the learning set elements . we will use this method in the next part of the paper . the quantum states obtained using the quantum quantization procedure can be used in any scheme for pattern recognition . for example , in the case of k - nn scheme , one needs to perform the following steps . * step :* perform quantum quantization by mapping class elements @xmath32 calculate the distances @xmath33 between the class representatives @xmath27 and the pattern for each class @xmath28 . classify the pattern according to the voting using values of @xmath34 uilizing @xmath35 values . however , in this case the procedure is reduced to the classical one as it does not utilize the superposition of the base vectors . for this reason we restrict our attention to the distance - base methods . in the following we focus on the features represented as numerical values . we start by considering the 1d feature vectors and introducing the quantum quantization procedure . in this case features ( and class elements ) are represented using flat representation in the subspace of the appropriate quantum state space . in this context flat means that patterns are represented by first transforming them into equally distributed intervals and next by representing the intervals as quantum state vectors . in orders to fix the dimension of the quantum state space required to encode the quantization symbols , we will use the data from the learning set . we also require that for the fixed state space of dimension @xmath36 , vectors @xmath37 and @xmath38 will encode all elements smaller and all elements larger than the elements in the learning set . for this reason we required that in the case of two classes @xmath39 - [ \min(c_1 \cup c_2)]\| + 2.\ ] ] here @xmath40 $ ] is the nearest integer function , sometimes denoted as @xmath41 , defined as @xmath42 = \left\ { \begin{array}{ll } -1^{\mathrm{sgn}(c ) } \lceil \|c\| \rceil , & \mathrm{if}\ \{c\}<1/2\\ -1^{\mathrm{sgn}(c ) } \lfloor \|c\| \rfloor , & \mathrm{if}\ \{c\}\geq 1/2 \end{array}\right.,\ ] ] where @xmath43 denotes the integer part of @xmath44 , latexmath:[$\{c\ } = \|\lfloor \|c\| \rfloor - standard library and as ` round ` function in _ mathematica _ system . for a given number @xmath44 we construct its quantum representation as vector @xmath46 - \{\min(c_1 \cup c_2)\ } + 1 \rangle}},\ ] ] in @xmath36-dimensional space of pure quantum states . the above procedure can be implemented in _ mathematica _ system as .... qmap[x _ , ns1 _ , ns2 _ ] : = block[{qmin , qmax , qdim } , qmin = round[min[join[ns1 , ns2 ] ] ] ; qmax = round[max[join[ns1 , ns2]]]+1 ; qdim = abs[qmax - qmin ] + 2 ; which [ x < qmin , ket[0 , qdim ] , x > qmax , ket[qdim - 1 , qdim ] , true , ket[round[x ] - qmin + 1 , qdim ] ] ] .... using the above quantization we define for each class its representation as @xmath47 where @xmath48 represents an unnormalized superposition of the elements representing the class . finally we introduce quantum flat subspace ( qfs ) classifier using the score function @xmath49 for @xmath50 representing classes @xmath51 . using this we classify pattern @xmath44 into class @xmath52 iff @xmath53 for all @xmath54 . let us consider two samples of real numbers @xmath55 and @xmath56 distributed with @xmath57 and @xmath58 respectively . for the purpose of classification we represent each class as a normalized superposition of its representatives . for example if @xmath55 is represented by @xmath59 and @xmath56 by @xmath60 we get quantum representations for both classes as @xmath61 and @xmath62 respectively for @xmath55 and @xmath56 . both @xmath63 and @xmath64 are normalized and each component has an amplitude representing a relative frequency of the appropriate element in the learning set . one should note that we need to choose the dimension of the subspace encoding the integer part to be large enough for encoding all representatives . for this reason the representation will require higher dimensional state space if the learning set contains more elements . in this example the observations are encoded in the space of dimension @xmath65 . this classification gives the following output @xmath66 where each pair represents a number and the assigned class . note that number 4 has been classified as belonging to class 1 due to the lack of the representative corresponding to this number in both classes . for this reason @xmath67 and the classifier can not decide about the result solely on the data included in the quantum state . to avoid the inconclusive results we assign the pattern to the arbitrarily chosen class . in this section we provide methods for using the quantum encoded patterns consisting of vectors of features . we introduce two basic schemes for encoding patterns and classes separable and non - separable . next , we provide some examples of using these schemes . we also propose a hybrid method which only partially exploits the tensor product structure of multipartite quantum states . let us use the learning set illustrated in fig [ fig : ent_vs_sep ] as an initial example . in this case the elements of the class are represented as quantum states @xmath68 and @xmath69 . we can use this representation and construct the representation of the class as the vector @xmath70 . in this situation we obtain the state which is entangled . however , we can alternatively construct the representation of the class in fig . [ fig : ent_vs_sep ] component - wise and in this case we obtain a pair of vectors resulting from the quantum averaging of the components , namely @xmath71 . this can be translated into the representation in the tensor product space as @xmath72 . the most straightforward extension of the 1d example to the case of patterns represented by vectors is obtained by using the 1d classifier independently from each component . in this model the features are encoded into corresponding quantum states and the classification is performed component - wise . as a result one obtains a vector of results , which indicates the classification of the pattern . following the example presented above , we introduce the separable data representation as follows . for a vector of features @xmath73 the quantum counterpart obtained using the separable data representation is constructed as @xmath74-tuple of quantum representations of the components , _ ie._@xmath75 accordingly , one can introduce the quantum averaging used to represent classes of features as an averaging over the components . this can be expressed as follows . for a class consisting of @xmath74-dimensional features vector , @xmath76 , the quantum separable representation of the class is obtained as @xmath77 where @xmath78 denotes the @xmath79-th element of the @xmath80-th feature vector in the class @xmath81 . in the simplest case of 2d feature space we have @xmath82 and the class representations are constructed as @xmath83 one should note that in the introduced scheme the dirac notation is overloaded and has to be interpreted according to the type of its argument . moreover , the quantum quantization procedure required to build the representations of components has to operate on the quantum state space of dimension appropriate for each feature . one of the main differences between the classical and quantum mechanics is the structure of the space used to describe the composite system . in the classical case such space is constructed using the cartesian product of the subspaces describing the subsystem , whilst in the quantum case the composite systems are described using the kronecker product . in quantum information theory the main phenomenon arising from the use of the tensor product is the entanglement . as the entanglement can be seen as a form of strongly non - classical correlation , we expect that the usage of the tensor product structure allows us encoding and exploiting the correlation between the features . for this reason we introduce the non - separable scheme as follows . for a vector of features @xmath73 the quantum counterpart obtained using the non - separable data representation is constructed as a tensor product of quantum representations of the components , _ ie._@xmath84 in the simplest case of 2d feature space we have @xmath85 for a class consisting of @xmath74-dimensional features vector , @xmath76 , the quantum non - separable representation of the class is obtained as @xmath86 the main difference between the above formula and the formula in eq . ( [ eqn : rep - class - separable ] ) is that it maps feature vectors onto vector space of dimension @xmath87 , where @xmath88 , @xmath50 , are dimensions of spaces required to encode features . the separable data representation requires only the space of dimension @xmath89 . in the case of 2d feature space the above encoding can be implemented in _ mathematica _ as .... qmuent2d[k_,ns _ ] : = block[{m1 , m2 , v } , m1 = qmap [ # , ns[1 , 1 ] , ns[2 , 1 ] ] & /@ ns[k , 1 ] ; m2 = qmap [ # , ns[1 , 2 ] , ns[2 , 2 ] ] & /@ ns[k , 2 ] ; v = plus @@ flatten[table[x $ \otimes$ y , { x , m1 } , { y , m2 } ] , 1 ] ; v / norm[v ] ] .... where function ` qmap ` was defined in section [ sec : representation - flat ] . the main advantage of the scheme introduced in this section is that it can be easily implemented and executed for the real - world data . in this section we present a simple example illustrating the difference between the separable and the non - separable representation of data . in order to execute the classification using the representations introduced above , one has to specify the score function used in the final step . in the case of the non - separable data representation we use the natural choice , namely the overlap between the class representation and the quantum state representing the pattern , @xmath90 in the case of the separable data representation we will use the score function defined as @xmath91 where @xmath92 denotes the @xmath35-th component of the separable quantum representation defined in eq . ( [ eqn : rep - class - separable ] ) . the above definitions allow the comparison of the efficiency of the pattern classification . for this purpose we have used two 2d gaussian sources producing the pairs described as @xmath93 and @xmath94 . for the purpose of the presented experiment we have generated a sample of 100 learning sets , each consisting of @xmath74 elements of both classes , with @xmath95 . for each learning set we have used two samples of 1000 points generated using the distribution describing the classes . each of the generated points has been classified using both separable and non - separable quantum data representation . the test has been assessed as positive if the point had been classified to the class corresponding to the appropriate distribution . the results of the described experiment are presented in fig . [ fig : succ - rates - test1 ] . as one can observe the utilization of the non - separable data representation allows better classification of the test data . this is expected as the utilization of the tensor product enables the operations on the larger subspace for the purpose of encoding the quantization symbols . in the presented work we have introduced the complete quantum - inspired scheme enabling the processing of real - world data for the purpose of pattern classification . the introduced representations of data are based on the quantization procedure which reduces the dimensionality of the quantum state space required to process the data . thanks to this one can implement and execute the introduced scheme on a standard desktop computer . we have also used the introduced framework to conduct a simple experiment where it is possible to observe the difference between the separable and non - separable representation of data . one should note that the presented considerations are based on the formalism of quantum mechanics , but do not require a working quantum computer . the utilization of a quantum machine would be beneficial from the security point of view as in this case the information about the class representatives is unknown to the party performing the classification . the introduced framework is based on a quantization procedure and the introduction of the linear combinations of quantization symbols . as such it can be easily adopted for other problems in machine learning . in particular we plan to apply it to the problem of clustering . from the quantum information point of view , the introduced framework can be generalized by using a formalism of mixed states in the quantization procedure . one of the advantages of such approach is the possibility of using other distance measures , based on the generalization of distance measures used in classical information theory . moreover , the use of density matrices can be used for incorporating the information about uncertainty in the measurement results . this work has been supported by ras project on : `` modeling the uncertainty : quantum theory and imaging processing '' , lr 7/8/2007 . the author would like to thank p. gawron , p. glomb , g. sergioli and l. didacci for helpful discussions . cunningham , p. , cord , m. , delany , s. : supervised learning . in : cord , m. , cunningham , p. ( eds . ) machine learning techniques for multimedia , pp . cognitive technologies , springer berlin heidelberg ( 2008 )	we provide a complete work - flow , based on the language of quantum information theory , suitable for processing data for the purpose of pattern recognition . the main advantage of the introduced scheme is that it can be easily implemented and applied to process real - world data using modest computation resources . at the same time it can be used to investigate the difference in the pattern recognition resulting from the utilization of the tensor product structure of the space of quantum states . we illustrate this difference by providing a simple example based on the classification of 2d data . * keywords : * pattern discrimination ; quantum computing ; entanglement of data
the study of heavy quarkonium production is particularly interesting because it may provide decisive information in understanding hadronization of heavy quarks and gluons in qcd . the most widely accepted theory to describe heavy quarkonium production at present is nonrelativistic qcd ( nrqcd ) factorization @xcite , in which the production is factorized into perturbative calculable short - distance coefficients and nonperturbative ( and universal ) long - distance matrix elements ( ldmes ) . as short - distance coefficients can be expanded in strong - coupling @xmath7 and each ldme has a definite power in @xmath8 ( the velocity of heavy quarks in the rest frame of heavy quarkonium ) , nrqcd factorization gives predictions by double expansion in @xmath7 and @xmath9 . nrqcd factorization is efficient such that in principle only a finite number of universal parameters , which can be determined by using some known experimental data , are involved with required precision in predicting other production processes . although a complete proof of factorization is still lacking , at least it holds up to next - to - next - to - leading order in @xmath7 @xcite . based on nrqcd factorization , charmonium production in hadron colliders has been studied extensively in recent years@xcite . specifically , for @xmath10 hadroproduction , it is found that all data at large @xmath11 , including both yield and polarization , can be well described by nrqcd factorization if one chooses a large @xmath12 and a small @xmath13 @xcite ( see also refs.@xcite ) . here @xmath12 and @xmath13 are linear combinations of related ldmes which are defined in refs.@xcite and will also be mentioned below , and roughly speaking , their values represent the importance of @xmath14 behavior and @xmath15 behavior in @xmath10 production cross sections , respectively . there are reasons that studying bottomonium may be a more suitable choice than charmonium to test the nrqcd factorization formalism . first , the value of @xmath9 is smaller in bottomonium ( @xmath16 ) than that in charmonium ( @xmath17 ) , thus the expansion in @xmath9 should converge faster in bottomonium . second , the mass of bottomonium is about 3 times of that of charmonium , then asymptotic freedom implies the convergence of @xmath18 expansion is also better in bottomonium . however , on the experiment side , the situation is not so satisfactory as the production rates of bottomonium are much smaller than charmonium , e.g. , the cross section of @xmath0 is about 2 orders of magnitude smaller than that of @xmath10 . furthermore , there are more excited bottomonium states which are below the open bottom ( say @xmath19 ) threshold and can decay into lower bottomonium states such as @xmath0 with large branching ratios and consequently contribute a substantial fraction to the lower bottomonium inclusive production by the so - called feed - down contributions , thus it is hard to measure the direct production from the prompt inclusive production . in the lhc era , we expect these disadvantages may be overcome by the higher luminosity , thus testing nrqcd factorization by bottomonium production seems to be hopeful . the inclusive differential cross section and polarization of @xmath20 are measured at the tevatron @xcite , but the @xmath0 polarizations observed by d0 @xcite and cdf @xcite disagree with each other . furthermore , both the d0 and cdf measurements contradict the lo nrqcd prediction @xcite . as argued in refs.@xcite , the next - to - next - to - leading order and even higher - order contributions @xcite may not be important , as compared with the full next - to - leading order ( nlo ) qcd contributions including both color - singlet ( cs ) and color - octet ( co ) channels , which are essential in understanding the @xmath20 ( and similarly the @xmath10 ) hadroproduction . partial nlo qcd contributions to @xmath20 hadroproduction have been calculated recently @xcite , and it is found that the nlo qcd corrections of s - wave co channels only slightly change the transverse momentum distribution and the polarization , while the correction of cs channel may bring on significant enhancement to the momentum distribution and change the polarization from transverse at lo into longitudinal at nlo . but the nlo contributions of p - wave channels for @xmath20 hadroproduction are still missing . at the lhc , cms has published the first run data for @xmath20 production @xcite , and lhcb has also reported the measured result @xcite . thus it is timely to present a complete nlo theoretical prediction for @xmath20 production , and compare theory with experiment . in this work , we study the @xmath0 hadroproduction in the framework of nrqcd , including all nlo contributions and feed - down contributions . the paper is organized as follows . we briefly introduce our calculation in sec . [ sec : method ] . then in sec . [ sec : feeddown ] , we describe our method for taking into account the feed - down contributions . ( note that the feed - down contributions for @xmath0 have not been treated seriously in all previous theoretical works . ) in sec . [ sec : result ] , we fit data to determine ldmes and then give predictions for the lhc experiment . finally , a summary is given in sec . [ sec : summary ] . the method of nlo calculation used in this work is similar to that used in @xmath10 and @xmath21 production @xcite . for completeness , we will sketch it in this section . according to the nrqcd factorization formalism , the inclusive cross section for direct bottomonium @xmath22 production in hadron - hadron collisions is expressed as @xmath23&=\sum\limits_{n}d\hat{\s}[(b\bar{b})_n]\dfrac{\moh}{m_b^{2l_n}}\\ & = \sum\limits_{i , j , n}\int dx_1 dx_2 g_{i / p}g_{j / p } \times d\hat{\s}[i+j\rightarrow ( b\bar{b})_n + x]\moh , \end{split}\ ] ] where @xmath24 is either a proton or an antiproton , the indices @xmath25 run over all the partonic species , and @xmath26 denote the color , spin and angular momentum ( @xmath27 ) of the intermediate @xmath28 states . in this work , we calculate the cross sections up to @xmath29 corrections , so that the intermediate states include @xmath30}_1 $ ] , @xmath31}_j$ ] , @xmath32 , @xmath33 and @xmath34 . note that our definition of cs ldmes @xmath35})\rangle}$ ] and @xmath36})\rangle}$ ] are different from that in ref . @xcite by a factor of @xmath37 . the calculation proceeds with three steps : calculating the parton level differential cross section @xmath38 $ ] , integrating over the phase space , and fitting the ldmes . nlo corrections for the parton level differential cross section include virtual corrections and real corrections . for virtual corrections , we use feynarts @xcite to generate feynman diagrams and amplitudes . we then calculate these thousands feynman diagrams analytically using our self - written mathematica code . finally , we output the simplified expression into c++ code . because the infrared divergence will appear when doing phase space integration for the real correction , we use the two cutoff phase space slicing method @xcite to isolate the divergence . the contributions from the singular phase space part are calculated analytically , while finite parts are calculated by using the berends - giele off - shell recursive relations @xcite . in the analytical calculation we have checked that all the ultraviolet ( uv ) and infrared ( ir ) singularities are canceled exactly . the uv divergences are removed by renormalization . the ir singularities arising from loop integration and phase space integration of the real correction partially cancel each other . the remaining ir singularities are absorbed into the proton parton - distribution functions and the nrqcd ldmes . the numerical integration over the phase space is handled by our self - written c++ codes , where we also use both qcdloop @xcite and looptools @xcite to calculate the scalar functions in the virtual corrections numerically . we verified that our results are independent of the two cuts introduced by the phase space slicing method . the method of fitting ldmes will be discussed in sec . [ sec : result ] . one difficulty in predicting @xmath20 production cross section is the treatment of feed - down contribution . there are several higher excited states that can decay into @xmath39 and they include : @xmath40 , @xmath41 , @xmath42 , @xmath43 , @xmath44 and @xmath45 . unfortunately , there are not enough data to determine ldmes of these higher excited states , therefore , it is hard to predict their feed - down contributions to @xmath39 . in fact , all previous predictions that were based on nrqcd factorization did not have a serious treatment of the feed - down contributions . the key point to deal with feed - down contribution is to determine the relation between momentum of higher excited states and momentum of @xmath39 . in ref . @xcite , we find a very good approximation that the ratio of two momenta is inversely proportional to the ratio of their masses . notice that the mass differences between these excited states and @xmath39 are of the order of @xmath46 and @xmath9 is very small in bottomonium , as a result , unlike the @xmath10 case , the momentum shift can be ignored when these excited states decay into @xmath39 . hence the production ldmes can be approximately combined into 6 independent ones : @xmath47})\rangle}=&{\langle\mathcal{o}^{{\upsilon(1s)}}(\bigl.^3\hspace{-1mm}s_1^{[1]})\rangle } + \sum_{n=2,3}{\langle\mathcal{o}^{{\upsilon(ns)}}(\bigl.^3\hspace{-1mm}s_1^{[1]})\rangle}\br({\upsilon(ns)}\rightarrow{\upsilon(1s)}),\\ { \langle\mathcal{o}^{}(\bigl.^3\hspace{-1mm}p_1^{[1]})\rangle}= & \sum_{n=1,2}{\langle\mathcal{o}^{{\chi_{b1}(np)}}(\bigl.^3\hspace{-1mm}p_1^{[1]})\rangle}\br({\chi_{b1}(np)}\rightarrow{\upsilon(1s)}),\\ { \langle\mathcal{o}^{}(\bigl.^3\hspace{-1mm}p_2^{[1]})\rangle}= & \sum_{n=1,2}{\langle\mathcal{o}^{{\chi_{b2}(np)}}(\bigl.^3\hspace{-1mm}p_2^{[1]})\rangle}\br({\chi_{b2}(np)}\rightarrow{\upsilon(1s)}),\\ { \langle\mathcal{o}^{}(\bigl.^3\hspace{-1mm}s_1^{[8]})\rangle}=&{\langle\mathcal{o}^{{\upsilon(1s)}}(\bigl.^3\hspace{-1mm}s_1^{[8]})\rangle } + \sum_{n=2,3}{\langle\mathcal{o}^{{\upsilon(ns)}}(\bigl.^3\hspace{-1mm}s_1^{[8]})\rangle}\br({\upsilon(ns)}\rightarrow{\upsilon(1s)})\\ & + \sum_{n=1,2}\sum_{j=1,2}{\langle\mathcal{o}^{{\chi_{bj}(np)}}(\bigl.^3\hspace{-1mm}s_1^{[8]})\rangle}\br({\chi_{bj}(np)}\rightarrow{\upsilon(1s)}),\\ { \langle\mathcal{o}^{}(\bigl.^1\hspace{-1mm}s_0^{[8]})\rangle}=&{\langle\mathcal{o}^{{\upsilon(1s)}}(\bigl.^1\hspace{-1mm}s_0^{[8]})\rangle } + \sum_{n=2,3}{\langle\mathcal{o}^{{\upsilon(ns)}}(\bigl.^1\hspace{-1mm}s_0^{[8]})\rangle}\br({\upsilon(ns)}\rightarrow{\upsilon(1s)}),\\ { \langle\mathcal{o}^{}(\bigl.^3\hspace{-1mm}p_0^{[8]})\rangle}=&{\langle\mathcal{o}^{{\upsilon(1s)}}(\bigl.^3\hspace{-1mm}p_0^{[8]})\rangle } + \sum_{n=2,3}{\langle\mathcal{o}^{{\upsilon(ns)}}(\bigl.^3\hspace{-1mm}p_0^{[8]})\rangle}\br({\upsilon(ns)}\rightarrow{\upsilon(1s ) } ) . \end{split}\ ] ] here the @xmath48 feed - down into @xmath0 is ignored due to the smallness of the transition branching ratios . the potential model results of wave functions and their derivatives at the origin can be chosen as@xcite @xmath49 and the cs ldmes can be estimated by @xmath50})\rangle}&=\frac{3}{4\pi}\lvert r_{{\upsilon(ns)}}(0 ) \rvert^{2},\\ { \langle\mathcal{o}^{{\chi_{bj}(np)}}(\bigl.^3\hspace{-1mm}p_j^{[1]})\rangle}&=\frac{3}{4\pi}\lvert r_{{\chi_{b}(np)}}^{\prime}(0 ) \rvert^{2}(2j+1 ) . \end{split}\ ] ] with the pdg data of branching ratios @xcite , we get @xmath51})\rangle}= & 1.81 \gev^3,\\ { \langle\mathcal{o}^{}(\bigl.^3\hspace{-1mm}p_1^{[1]})\rangle}= & 0.54 \gev^5,\\ { \langle\mathcal{o}^{}(\bigl.^3\hspace{-1mm}p_2^{[1]})\rangle}= & 0.62 \gev^5 . \end{split}\ ] ] now there leave only 3 unknown co ldmes : @xmath4})\rangle}$ ] , @xmath5})\rangle}$ ] , and @xmath6})\rangle}$ ] . they will be determined by fitting the tevatron data @xcite . because they are nearly universal ( up to a correction of order @xmath9 with calculated short - distance coefficients in the fit ) , the fitted results can be used to predict @xmath39 production in other colliders . the cteq6l1 and cteq6 m parton - distribution functions @xcite are used for lo and nlo calculations , respectively . the bottom quark mass is set to be @xmath52 , while the renormalization , factorization , and nrqcd scales are @xmath53 and @xmath54 , where @xmath55 is the @xmath20 transverse mass . the center - of - mass energies are 1.8 @xmath56 , 1.96 @xmath56 and 7 @xmath56 for the tevatron run i , run ii , and lhc , respectively . [ cols="^,^ " , ] in summary , we calculate the complete nlo corrections for the @xmath0 production at hadron colliders up to @xmath57 . ignoring corrections of higher - orders in @xmath9 , we combine the production ldmes of @xmath0 and other excited states into 3 color - singlet ldmes and 3 color - octet ldmes . these 6 ldmes are approximately universal and they include almost all feed - down contributions to @xmath0 production . the cs ldmes are estimated by using potential model results , while the co ldmes are determined by fitting the tevatron data . then we find our predictions well coincide with the new experimental data at the lhc . our work may provide a new test for the universality of ldmes in @xmath0 hadroproduction . to have a comprehensive understanding of @xmath0 hadroproduction , it is certainly important to also compare the theoretical result with the polarization data for @xmath0 , we leave it as a further study . encouraged by the result of @xmath10 polarization @xcite , where we find the @xmath10 polarization and yield can be consistently explained by two well constrained co ldmes ( @xmath12 and @xmath13 ) , a good description for the @xmath0 data including yield and polarization seems to be promising . however , note that the values of two well constrained co ldmes in @xmath10 production are significantly different @xcite , while for @xmath0 production we find from eq . that @xmath58 and @xmath59 are of the same order . another complexity concerns the influence of big feed - down contributions on @xmath0 polarization . therefore , a full understanding of @xmath0 production including both yield and polarization may provide important information in addition to the study of @xmath10 production . on the experiment side , because the bottom quark is heavy : @xmath60 5 gev , to test the large @xmath11 ( @xmath61 ) behavior one needs to measure the cross sections and polarizations at @xmath11 as large as , say 30 gev and even larger , with higher statistics , and to separate the higher excited @xmath3 production from the @xmath0 production . this is a hard task for experiment , and we hope it can be fulfilled at the lhc in the near future . then we can make more thorough comparison between theory and experiment , and provide a further test of nrqcd factorization . this work was supported by the national natural science foundation of china under grant nos . 10721063 , 11021092 , 11075002 and the ministry of science and technology of china under grant no.2009cb825200 . g. t. bodwin , e. braaten , and g. p. lepage , `` rigorous qcd analysis of inclusive annihilation and production of heavy quarkonium , '' http://dx.doi.org/10.1103/physrevd.51.1125[_phys . * d51 * ( 1995 ) 1125 [ erratum ibid . d * 55 * ( 1997 ) 5853 ] ] , http://arxiv.org/abs/hep-ph/9407339[arxiv:hep-ph/9407339 ] . g. c. nayak , j .- w . qiu , and g. f. sterman , `` fragmentation , factorization and infrared poles in heavy quarkonium production , '' http://dx.doi.org/10.1016/j.physletb.2005.03.031[_phys . lett . _ * b613 * ( 2005 ) 45 ] , http://arxiv.org/abs/hep-ph/0501235[arxiv:hep-ph/0501235 ] . g. c. nayak , j .- w . qiu , and g. f. sterman , `` fragmentation , nrqcd and nnlo factorization analysis in heavy quarkonium production , '' http://dx.doi.org/10.1103/physrevd.72.114012[_phys . rev . _ * d72 * ( 2005 ) 114012 ] , http://arxiv.org/abs/hep-ph/0509021[arxiv:hep-ph/0509021 ] . p. artoisenet , j. p. lansberg , and f. maltoni , `` hadroproduction of @xmath62 and @xmath20 in association with a heavy - quark pair , '' http://dx.doi.org/10.1016/j.physletb.2007.04.031[_phys . * b653 * ( 2007 ) 6066 ] , http://arxiv.org/abs/hep-ph/0703129[arxiv:hep-ph/0703129 ] . he , r. li , and j .- x . wang , `` qed contribution to the production of @xmath63 at the tevatron and lhc , '' http://dx.doi.org/10.1103/physrevd.79.094003[_phys . rev . _ * d79 * ( 2009 ) 094003 ] , http://arxiv.org/abs/0904.2069[arxiv:0904.2069 [ hep - ph ] ] . d. li , y .- q . ma , and k .- chao , `` @xmath64 production associated with a @xmath65 pair at hadron colliders , '' http://dx.doi.org/10.1103/physrevd.83.114037[_phys.rev._ * d83 * ( 2011 ) 114037 ] , http://arxiv.org/abs/1106.4262[arxiv:1106.4262 [ hep - ph ] ] . j. m. campbell , f. maltoni , and f. tramontano , `` qcd corrections to @xmath62 and @xmath20 production at hadron colliders , '' http://dx.doi.org/10.1103/physrevlett.98.252002[_phys . lett . _ * 98 * ( 2007 ) 252002 ] , http://arxiv.org/abs/hep-ph/0703113[arxiv:hep-ph/0703113 ] . b. gong and j .- x . wang , `` next - to - leading - order qcd corrections to @xmath62 polarization at tevatron and large - hadron - collider energies , '' http://dx.doi.org/10.1103/physrevlett.100.232001[_phys . _ * 100 * ( 2008 ) 232001 ] , http://arxiv.org/abs/0802.3727[arxiv:0802.3727 [ hep - ph ] ] . b. gong and j .- x . wang , `` qcd corrections to polarization of @xmath62 and @xmath20 at tevatron and lhc , '' http://dx.doi.org/10.1103/physrevd.78.074011[_phys . rev . _ * d78 * ( 2008 ) 074011 ] , http://arxiv.org/abs/0805.2469[arxiv:0805.2469 [ hep - ph ] ] . b. gong , x. q. li , and j .- x . wang , `` qcd corrections to @xmath62 production via color octet states at tevatron and lhc , '' http://dx.doi.org/10.1016/j.physletb.2009.02.026[_phys . lett . _ * b673 * ( 2009 ) 197 [ erratum ibid . * 693 * , 612 ( 2010 ) ] ] , http://arxiv.org/abs/0805.4751[arxiv:0805.4751 [ hep - ph ] ] . y .- ma , k. wang , and k .- chao , `` qcd radiative corrections to @xmath64 production at hadron colliders , '' http://dx.doi.org/10.1103/physrevd.83.111503[_phys . rev . _ * d83 * ( 2011 ) 111503(r ) ] , http://arxiv.org/abs/1002.3987[arxiv:1002.3987 [ hep - ph ] ] . ma , k. wang , and k .- chao , `` @xmath66 production at the tevatron and lhc at @xmath67 in nonrelativistic qcd , '' http://dx.doi.org/10.1103/physrevlett.106.042002[_phys . * 106 * ( 2011 ) 042002 ] , http://arxiv.org/abs/1009.3655[arxiv:1009.3655 [ hep - ph ] ] . m. butenschen and b. a. kniehl , `` complete next - to - leading - order corrections to @xmath62 photoproduction in nonrelativistic quantum chromodynamics , '' http://dx.doi.org/10.1103/physrevlett.104.072001[_phys . * 104 * ( 2010 ) 072001 ] , http://arxiv.org/abs/0909.2798[arxiv:0909.2798 [ hep - ph ] ] . y .- ma , k. wang , and k .- chao , `` complete next - to - leading order calculation of the @xmath62 and @xmath68 production at hadron colliders , '' http://dx.doi.org/10.1103/physrevd.84.114001[_phys.rev._ * d84 * ( 2011 ) 114001 ] , http://arxiv.org/abs/1012.1030[arxiv:1012.1030 [ hep - ph ] ] . m. butenschoen and b. a. kniehl , `` j / psi polarization at tevatron and lhc : nonrelativistic - qcd factorization at the crossroads , '' http://dx.doi.org/10.1103/physrevlett.108.172002[_phys . rev . * 108 * ( 2012 ) 172002 ] , http://arxiv.org/abs/1201.1872[arxiv:1201.1872 [ hep - ph ] ] . chao , y .- q . ma , h .- s . shao , k. wang , and y .- j . zhang , `` @xmath62 polarization at hadron colliders in nonrelativistic qcd , '' http://link.aps.org/doi/10.1103/physrevlett.108.242004[_phys . * 108 * ( 2012 ) 242004 ] , http://arxiv.org/abs/1201.2675[arxiv:1201.2675 [ hep - ph ] ] . z .- kang , j .- w . qiu , and g. sterman , `` heavy quarkonium production and polarization , '' http://dx.doi.org/10.1103/physrevlett.108.102002[_phys . _ * 108 * ( 2012 ) 102002 ] , http://arxiv.org/abs/1109.1520[arxiv:1109.1520 [ hep - ph ] ] . d. e. acosta _ ( cdf collaboration ) , `` @xmath20 production and polarization in @xmath69 collisions at @xmath70 1.8-tev , '' http://dx.doi.org/10.1103/physrevlett.88.161802[_phys . * 88 * ( 2002 ) 161802 ] . v. m. abazov _ ( d0 collaboration ) , `` measurement of inclusive differential cross sections for @xmath0 production in @xmath71 collisions at @xmath72 = 1.96-tev , '' http://dx.doi.org/10.1103/physrevlett.94.232001[_phys . * 94 * ( 2005 ) 232001 ] , http://arxiv.org/abs/hep-ex/0502030[arxiv:hep-ex/0502030 ] . v. m. abazov _ ( d0 collaboration ) , `` measurement of the polarization of the @xmath73 and @xmath74 states in @xmath75 collisions at @xmath72 = 1.96-tev , '' http://dx.doi.org/10.1103/physrevlett.101.182004[_phys . lett . _ * 101 * ( 2008 ) 182004 ] , http://arxiv.org/abs/0804.2799[arxiv:0804.2799 [ hep - ex ] ] . t. kuhr ( f. t. c. collaboration ) , `` upsilon polarization measurement at cdf , '' _ pos _ * dis2010 * ( 2010 ) 159 , http://arxiv.org/abs/1011.0161[arxiv:1011.0161 [ hep - ex ] ] . t. aaltonen _ et al . _ ( cdf collaboration ) , `` measurements of angular distributions of muons from @xmath20 meson decays in @xmath69 collisions at @xmath76 tev , '' http://dx.doi.org/10.1103/physrevlett.108.151802[_phys . rev . * 108 * ( 2012 ) 151802 ] , http://arxiv.org/abs/1112.1591[arxiv:1112.1591 [ hep - ex ] ] . e. braaten and j. lee , `` polarization of @xmath77 at the tevatron , '' http://dx.doi.org/10.1103/physrevd.63.071501[_phys . _ * d63 * ( 2001 ) 071501 ] , http://arxiv.org/abs/hep-ph/0012244[arxiv:hep-ph/0012244 ] . p. artoisenet , j. m. campbell , j. p. lansberg , f. maltoni , and f. tramontano , `` upsilon production at fermilab tevatron and lhc energies , '' http://dx.doi.org/10.1103/physrevlett.101.152001[_phys . * 101 * ( 2008 ) 152001 ] , http://arxiv.org/abs/0806.3282[arxiv:0806.3282 [ hep - ph ] ] . b. gong , j .- x . wang , and h .- f . zhang , `` qcd corrections to @xmath20 production via color - octet states at the tevatron and lhc , '' http://dx.doi.org/10.1103/physrevd.83.114021[_phys.rev._ * d83 * ( 2011 ) 114021 ] , http://arxiv.org/abs/1009.3839[arxiv:1009.3839 [ hep - ph ] ] . v. khachatryan _ et al . _ ( cms collaboration ) , `` measurement of the inclusive upsilon production cross section in pp collisions at sqrt(s)=7 tev , '' http://dx.doi.org/10.1103/physrevd.83.112004[_phys.rev._ * d83 * ( 2011 ) 112004 ] , http://arxiv.org/abs/1012.5545[arxiv:1012.5545 [ hep - ex ] ] . r. aaij _ et al . _ ( lhcb collaboration ) , `` measurement of upsilon production in pp collisions at sqrt(s ) = 7 tev , '' http://arxiv.org/abs/1202.6579[arxiv:1202.6579 [ hep - ex ] ] . t. hahn , `` generating feynman diagrams and amplitudes with feynarts 3 , '' http://dx.doi.org/10.1016/s0010-4655(01)00290-9[_comput _ * 140 * ( 2001 ) 418 ] , http://arxiv.org/abs/hep-ph/0012260[arxiv:hep-ph/0012260 ] . b. w. harris and j. f. owens , `` the two cutoff phase space slicing method , '' http://dx.doi.org/10.1103/physrevd.65.094032[_phys . _ * d65 * ( 2002 ) 094032 ] , http://arxiv.org/abs/hep-ph/0102128[arxiv:hep-ph/0102128 ] . f. a. berends and w. t. giele , `` recursive calculations for processes with n gluons , '' http://dx.doi.org/10.1016/0550-3213(88)90442-7[_nucl . * b306 * ( 1988 ) 759 ] . r. k. ellis and g. zanderighi , `` scalar one - loop integrals for qcd , '' http://dx.doi.org/10.1088/1126-6708/2008/02/002[_jhep_ * 02 * ( 2008 ) 002 ] , http://arxiv.org/abs/0712.1851[arxiv:0712.1851 [ hep - ph ] ] . t. hahn and m. perez - victoria , `` automatized one - loop calculations in four and d dimensions , '' [ _ comput . * 118 * ( 1999 ) 153 ] , http://arxiv.org/abs/hep-ph/9807565[arxiv:hep-ph/9807565 ] . e. j. eichten and c. quigg , `` quarkonium wave functions at the origin , '' http://dx.doi.org/10.1103/physrevd.52.1726[_phys . rev . _ * d52 * ( 1995 ) 1726 ] , http://arxiv.org/abs/hep-ph/9503356[arxiv:hep-ph/9503356 ] . collaboration , k. nakamura _ et al . _ , `` review of particle physics , '' http://dx.doi.org/10.1088/0954-3899/37/7a/075021[_j . _ * g37 * ( 2010 ) 075021 ] . m. r. whalley , d. bourilkov , and r. c. group , `` the les houches accord pdfs ( lhapdf ) and lhaglue , '' http://arxiv.org/abs/hep-ph/0508110[arxiv:hep-ph/0508110 ] .	with nonrelativistic qcd factorization , we calculate the @xmath0 prompt production at hadron colliders at next - to - leading order in @xmath1 . in addition to the color - singlet contribution , color - octet channels ( especially the p - wave channel ) up to @xmath2 are all considered . aside from direct production , the feed - down contributions from higher excited s - wave and p - wave @xmath3 states to @xmath0 production are also included . we use the potential model estimates as input for color - singlet long - distance matrix elements ( ldmes ) . while for color - octet contributions , we find they can be approximately described by three ldmes : @xmath4})\rangle}$ ] , @xmath5})\rangle}$ ] and @xmath6})\rangle}$ ] . by fitting the tevatron data we can determine some linear combinations of these ldmes , and then use them to predict @xmath0 production at the lhc . our predictions are consistent with the new experimental data of cms and lhcb .
recent _ chandra _ and _ xmm - newton _ observations of cooling flow ( cf ) clusters of galaxies have lead to a shower of papers examining reheating processes of the intracluster medium ( icm ) , which together with pre-_chandra _ and xmm - newton reheating papers make a long list of papers with different ideas ( e.g. , binney & tabor 1995 ; tucker & david 1997 ; ciotti & ostriker 2001 ; david et al . 2001 ; quilis , bower , & balogh 2001 ; brggen , m. & kaiser 2001 ; ruszkowski & begelman 2002 ; nulsen et al . 2002 ; fabian 2003 ; fabian et al . 2003 ; binney 2003 ; mathews et al . 2003 ; for more references see peterson et al . these recent observations , although in contradiction ( fabian 2003 ) with the cluster - cf models of 10 years ago ( fabian 1994 ) , are compatible with models having low mass cooling rates ( see binney 2003 , for why the new results are not a surprise ) , such as in the moderate cf model ( soker et al . the moderate cf model , which was proposed before the new results from the two x - ray telescopes , is different from many early proposed processes which aim is to prevent the cf in cluster of galaxies altogether . the main ingredient of the moderate cf model is that the effective age , i.e. , the time period since the last major disturbance of the icm inside the cooling radius @xmath2 , is much shorter than the cluster age ( e.g. , binney & tabor 1995 ; see binney 2003 ; soker et al . the cooling radius is defined as the place where the radiative cooling time equals the cluster age . originally the heating in the moderate cf model was proposed to be intermittent ( soker et al . 2001 ) , but the basic idea can hold for a steady heating , or heating in short intervals ( binney 2003 ) . kahn et al . ( 2003 ) and peterson et al . ( 2003a ) , among others , find that there is less gas than that predicted by old versions of the steady - state isobaric cf model at all temperatures below the ambient gas temperature . also , the discrepancy increases with decreasing temperature . there is not just a deficit in gas below @xmath3 of the ambient cluster temperature , there is a deficit of gas at all temperatures below the ambient temperature ( peterson et al . soker & david ( 2003 ) show that these xmm - newton results are consistent with the expectations of the moderate cf model . soker & david ( 2003 ) claim that the x - ray emitting gas in cf clusters can only be in steady - state if there exists a steady heating mechanism that scales as @xmath4 where @xmath5 ; this is unlikely . they show then that in the moderate cf model most of the gas within cf resides in the hottest gas , which is prevented from cooling continuously and attaining a steady - state configuration by being reheated . this results in a mass cooling rate that decreases with decreasing temperature , with a much lower mass cooling rate at the lowest temperatures , as discovered by peterson et al . ( 2003a ) . although there is no indication yet for a violent heating , i note the following . an intermittent reheating can work for an agn kinetic energy of @xmath6 , and its strong activity should last @xmath7 yr and occur every @xmath8 yr ( soker et al . a violent heating event by agn jets was proposed by kawano , ohto , & fukazawa ( 2003 ) to have occurred @xmath9 yr ago in the cf clusters 2a 0335 + 096 and a 2199 . mazzotta , edge , & markevitch ( 2003 ) claim that the structure of the cluster 2a 0335 + 096 can be accounted for if strong agn activity periods are followed by relative quiescence periods . osullivan & vrtilek ( 2003 ) propose that a heating event occured @xmath10 ago in the poor cluster mkw 4 , which consequently is presently not in a steady state cooling . such a scenario is in accord with the explanation of soker & david ( 2003 ) to the observations of peterson et al . ( 2003a ) and kahn et al . only @xmath11 of all cf clusters should be found during that stage . in addition , heinz & enlin ( 2003 ) argue for bias against detecting strong shocks . therefore , from the well studied cf clusters , only one , or at most two , are expected to be in the stage of violent heating . in cygnus a there is a strong radio source which heats the icm ( smith et al . wilson , young , & smith ( 2003 ) estimate the mechanical power of the jets in cygnus a to be @xmath12 . this is @xmath13 times larger than the radio emission , and it is in the range required by the moderate cf model . therefore , present observations do not rule - out a violent intermittent reheating process acting for a short time every @xmath8 yr . the limit on the cooling rate below a temperature @xmath14 inferred from x - ray observations is @xmath15 of the mass cooling rates cited in the past ( fabian 2002 ) . that is , x - ray emission from gas below @xmath14 is absent in present x - ray telescopes . this minimum temperature is different for different clusters , but it is @xmath3 of the background temperature ( kaastra et al . 2003 ; peterson et al . despite the missing x - ray emission reported by these studies , there are other indications that cooling to temperatures much below @xmath14 does occur , but on a much lower rate than that inferred by taking the cf age to be equal to the cluster age . wise , mcnamara , & murray ( 2003 ) and mcnamara , wise , & murray ( 2003 ) , for example , find the cooling rate within @xmath16 of the cf cluster a 1068 to be about equal to the star formation rate there @xmath17 . although the limits on the mass cooling rates cited above are within the values expected in the moderate cf model , a much better agreement with observations can be achieved if some non - radiative cooling of the x - ray gas is added to the moderate cf model ( soker , blanton , & sarazin 2003 , hereafter sbs ) . the thermal energy released in cooling the hot gas is delivered to the cold clouds , and is radiated as optical / uv line emission at @xmath18 k. this energy of the hot gas can be delivered to the cool gas by mixing ( e.g. , oegerle et al . 2001 ; fabian et al . 2001 , 2002 ; johnstone et al . 2002 ; bayer - kim et al . 200 ) , or via heat conduction ( sbs ) kpc ) heat conduction models aiming at heating , rather than cooling , the x - ray emitting gas in the center of cf clusters . ] in the scenario proposed by sbs the heat conduction is regulated by reconnection between the magnetic field lines in cold ( @xmath19 k ) clouds and the field lines in the icm . a narrow conduction front is formed , which , despite the relatively low temperature , allows efficient heat conduction from the hot icm to the cold clouds . the reconnection between the field lines in cold clouds and those in the icm occurs only when the magnetic field in the icm is strong enough . this occurs only in the very inner regions of cf clusters , at @xmath20 kpc . while in the scenario of sbs the non - radiative cooling starts with the reconnection of magnetic field lines from the pre - existing cold clouds to the x - ray emitting gas , in the present paper i examine heat conduction within pre - existing magnetic flux loops of x - ray emitting gas . the non - radiative cooling starts when small portion of the loop is radiatively cooling to @xmath21 . magnetic flux loops in cf clusters were studied before , e.g. , in relation to global heat conduction ( norman & meiksin 1996 ) , in the amplification of magnetic fields in the center of cf clusters ( godon , soker , & white 1998 ) , and the formation of multiphase icm ( zoabi , soker , & regev 1996 , 1998 ; hereafter zsr ) . in a recent paper kaastra et al . ( 2003 ) note the similarity of the differential emission measure in the clusters they study to that of coronal loops in the sun . the behavior of loops found by zsr , the finding of kaastra et al . ( 2003 ) , and the dynamically important magnetic field in the inner regions of cluster cfs ( eilek & owen 2002 ; taylor , fabian , & allen 2002 ) , motivate me to study non - radiative cooling within magnetic flux loops . this paper is not aiming at completely suppressing cf , as there are many evidence for cooling to low temperatures , e.g. , kaastra et al . ( 2003 ) find radiation from gas at temperatures of @xmath22 in some clusters , edge et al . ( 2002 ; see also edge & frayer 2003 ; salome & combes 2003a , b ; edwards 2003 ) find massive cold reservoir in the form of molecular gas , probably from cooling icm , and the star formation rate cited above ( wise et al . the main goal is to study another aspect , i.e. , non radiative cooling of x - ray emitting gas in magnetic flux loops , and incorporate it into the moderate cluster cf model . as such , this paper is complementary to sbs . zsr consider an elliptically shaped magnetic flux loop moving radially inward in a cluster cf . they follow only the two end points of the loop , assuming that they evolve in series of hydrostatic equilibria , i.e. , magnetic tension @xmath23 balances buoyancy @xmath24 at each of the two end points of the loop . this dictates that the density in the upper segment of the loop @xmath25 , is lower than that of its surroundings , and the density of the lower segment of the loop @xmath26 , is higher than that of its surroundings . zsr consider radiative cooling , but neglect heat conduction , even along the magnetic field lines . they base this assumption on a possible reduction in the heat conduction coefficient along magnetic fields lines ( pistinner & eichler 1998 ) . this reduction in heat conduction is not efficient when the magnetic pressure becomes comparable to the thermal pressure ( pistinner & eichler 1998 ) . zsr take the magnetic pressure to increase as density increases , and take a pressure equilibrium between the loop and its surroundings . as the loop evolves and falls toward the cluster center it cools with the lower - segment temperature @xmath27 falling faster than the upper - segment temperature @xmath28 . the loop is stretched in the radial direction and gets narrow on its sides ( see fig . zsr find that because the magnetic tension prevents the dense lower segment of the loop from falling toward the cluster center , this lower segment reaches a phase of catastrophic cooling . for the parameters used by zsr the catastrophic cooling occurs when the lower segment is inside a cluster radius of @xmath29 , but the upper regions of the loop extend to much larger distances , depending on its size . in the present paper i examine possible effects of heat conduction along magnetic field lines inside the loop . the two relevant time - scales are the conduction time scale along the loop , from the upper hot segment to the cooling lower segment , and the radiative cooling time at the denser lower segment . heat conduction proceeds on a time scale of @xmath30 , where the first parenthesis is the enthalpy per unit volume , and the second parenthesis is the heat loss per unit volume per unit time because of heat conduction . the temperature @xmath31 is some average inside the loop , which at early stages is close to both @xmath28 and @xmath27 . before the lower segment of the flux loop rapidly cools to @xmath32 , the temperature gradient inside the loop is shallower than @xmath33 , and the heat conduction time is longer than the one given above . here , @xmath34 is ( nearly ) a constant defined such that @xmath35 is the heat conduction coefficient , @xmath36 is the total number density , and @xmath37 the width of the conduction front , taken here to be the length of the magnetic flux loop ( see figure 1 ) . 0.2 cm scaling with typical values gives @xmath38 where @xmath39 is the electron density . at early stages , the time @xmath40 is the time required to transfer heat from the upper segment to the lower one to prevent it from cooling to very low temperatures . the isobaric radiative cooling time of the lower segment in the temperature range @xmath41 k is ( see cooling curve in , e.g. , gaetz , edgar , & chevalier 1988 ) @xmath42 the heat conduction will not prevent the lower segment from cooling to lower temperatures when the condition @xmath43 is met . using equations ( 1 ) and ( 2 ) , and substituting @xmath27 for @xmath31 ( this is reasonable before the catastrophic cooling of the lower segment starts ) , the condition for rapid cooling of the lower segment reads @xmath44 where i took a constant pressure ( expressed as @xmath45 , and scaled with @xmath46 ) , such that the condition is on the temperature of the lower loop s segment . it is evidence that the temperature when rapid cooling starts depends weakly on the external pressure and on the loop size . for the typical parameters used here , it is in the range of the minimum temperature of x - ray emitting gas deduced from observations ( peterson et al . 2003a ; kaastra et al . 2003 ) . to summarize this section , there are two conditions for the catastrophic cooling of the lower segment of a flux loop to occur . first , the magnetic field must be strong , close to equipartition , in order to prevent the dense segment from falling in too fast . this condition is also true for the scenario of sbs , where the strong magnetic field is required to initiate reconnection between cold clouds and the icm . second , the lower segment temperature must reach the value given by equation ( 3 ) . both conditions are likely to be met only near the cluster s center . borkowski , balbus , & fristrom ( 1990 ) numerically study the evolution of a conduction front between a gas at @xmath47 and cooler gas at @xmath48 . they start with a sharp boundary between the two media . after a short period of time , during which the conduction front widens , a quasi - static front is developed . in the third phase@xmath49the condensation phase@xmath49the entire front radiatively cools down . during the quasi - static front phase borkowski ( 1990 ) find the mean temperature of different ions to be @xmath50 , i.e. , much below the temperature of their hot phase which is @xmath51 . i consider a similar effect here , where the mean temperature is low , with most radiation emitted at much lower temperatures than @xmath28 . the initial situation here is not a sharp boundary between the hot and cold gases , but rather the lower segment of the loop cools from a high temperature . therefore , i expect the quasi - static front phase to be established quite early . an accurate solution for the temperature distribution along a static magnetic flux tube was derived analytically by rosner , tucker , & vaiana ( 1978 ) and numerically by aschwanden & schrijver ( 2002 ) . both works consider heating as well as heat conduction and radiative cooling . kaastra et al . ( 2003 ) when comparing their spectra to the results of aschwanden & schrijver ( 2002 ) assume the existence of a heating source , but they do nt specify its nature . i here neglect heating , because i assume sporadic heating events in the cluster , which most loops do not encounter during their catastrophic cooling . the results of rosner et al . ( 1978 ) can be reasonably fitted along most of the tube by @xmath52 with @xmath53 , where @xmath54 is the coordinate along the flux tube , i.e. , along the magnetic field lines , and @xmath55 is the location of the already cold gas ; here it is the lower segment of the loop . only at the highest temperature segment the temperature pofile becomes shallower . practically the temperature of the lower segment after its catastrophic cooling can be set to @xmath56 , although it is @xmath21 . the numerical results of aschwanden & schrijver ( 2002 ) depend on the assumptions used , but are crudely fit by @xmath57 . these results fit the quasi - static heat conduction phase discussed here , as i now show . a simple balance between radiative cooling and heat conduction at each point along the flux loop reads ( in the temperature range @xmath41 ) @xmath58 where @xmath59 is the cooling function , with @xmath60 in the relevant temperature range . to obtain a simple analytical solution i take a constant pressure along the flux tube , such that @xmath61 . this allows the last equation to be written as @xmath62 where@xmath63 is a constant composed of the different coefficients in equation ( 4 ) . this can be integrated once to yield @xmath64 where @xmath65 is a constant of integration to be determined by boundary conditions . the left hand side of the last equation is the heat flux . rosner et al . ( 1978 ) take this flux to be equal to zero at the higher temperature end of the loop . however , we see that such a boundary condition here requires @xmath66 , which is not possible for lower temperatures . a constant heating along the flux tube can make this solution possible , by balancing the slow cooling of the high temperature ( and low density ) segment . otherwise heat conduction balances radiative cooling and the heat flux ca nt vanish , and on the hottest end of the loop no real steady state situation exists . equation ( 4 ) can analytically be integrated in principle . however , the expression gives @xmath54 as function of @xmath31 and it is too complicated for practical uses . instead i look for a solution in the form @xmath52 . i start by taking a minimum heat flux , which vanishes when @xmath56 . this gives @xmath67 . the solution reads ( with @xmath55 where @xmath56 ) @xmath68 where @xmath69 . as this is close to the solution obtained by rosner et al . ( 1978 ) , i consider it a satisfactory approximation . @xmath70 gives a lower value of @xmath36 . for the extreme case of @xmath71 in the entire range ( which implies a huge heat flux , hence it is not appropriate here where no heating is considered ) , the solution gives @xmath72 . the variation of @xmath31 with @xmath54 is somewhere between these two extremes , i.e. , @xmath73 , but very likely close to @xmath74 . for a constant pressure along the loop , the relative emission per unit volume is given by @xmath75 . for a temperature given by @xmath52 , and a constant pressure along the loop , the luminosity fraction of gas at temperature @xmath76 is given by ( assuming constant loop s cross section ) @xmath77 where in the second equality i substituted @xmath78 , and preformed the integration . note that @xmath28 is not the maximum temperature in the cf cluster @xmath79 , but rather the maximum temperature inside the specific magnetic flux loop under consideration , and @xmath80 . for @xmath53 the luminosity of gas at temperature @xmath76 goes as @xmath81 . lower values of @xmath36 result in more emission from high temperature gas , but the higher value of @xmath82 seems appropriate for the following reason . i assumed above that the pressure along the flux tube is constant . however , in clusters the upper segment of the loop will be further out in a lower pressure environment . outside @xmath83 in a 2052 ( blanton et al . 2001 ) and @xmath84 in hydra a ( david et al . 2001 ) , the pressure drops as @xmath85 . this implies a pressure drop by a factor of @xmath86 between the two ends of a loop with @xmath87 . the lower pressure , hence density , at the upper loop segments implies less emission from the hot gas compared to the assumed constant pressure case . therefore , taking the higher value of @xmath36 in the range given above is justified . for the @xmath53 case , hence @xmath88 , @xmath89 of the luminosity comes from gas at @xmath90 , and @xmath91 from , gas at @xmath92 . for example , if @xmath93 ( recall that @xmath28 is not the maximum temperature in the cf cluster ) , then @xmath89 ( @xmath91 ) of the luminosity comes from gas at temperatures @xmath94 ( @xmath95 ) . overall , heat conduction within a magnetic flux loop can reduce the fraction of energy radiated in the x - ray band by a factor of @xmath0 . i found above that heat conduction within a magnetic flux loop can reduce the fraction of energy radiated in the x - ray band by a factor of @xmath0 , i.e. , the x - ray luminosity is @xmath96 of the value expected without heat conduction . cooling of x - ray emitting gas by conduction to cold gas was proposed to occur via mixing ( e.g. , oegerle et al . 2001 ; fabian et al . 2001 , 2002 ; johnstone et al . 2002 ; bayer - kim et al . 200 ) and by local heat conduction regulated by magnetic field reconnection ( sbs ) . in the later two processes the temperature gradient is much higher than the one inside the loop studied here , and hence these mechanism are more efficient in reducing x - ray emission . however , the question is whether these processes occur , and how often . the magnetic flux loop model studied here is motivated by previous suggestions that many such loops exist in the icm of cf clusters and that they play an active role in cooling and/or dynamics of the gas ( e.g. , zoabi et al . 1996 ; zsb ; kaastra et al . 2003 ) . sbs estimate that in their heat conduction scenario via magnetic field reconnection the total line emission expected for gas cooling from @xmath97 k to @xmath98 k might be factor of @xmath99 lower than that expected in older versions of the cf model , where only line emission is considered for the cooling process , and where the cooling rates are much higher . this factor comes from a @xmath1 times lower mass cooling rate of the hot gas in the moderate cf model ( soker et al . 2001 ) , and from another factor of @xmath100 for cooling via heat conduction . this much lower expected radiative cooling rate at low x - ray emitting temperatures is compatible with limits inferred from observations . the observational limit , which is @xmath101 of the expected cooling rate in old versions of the cf model ( fabian 2002 ; peterson et al . 2003a ) , can be met if the reduction due to heat conduction is only a factor of @xmath102 as found here , with another factor of @xmath1 from the moderate cf model . possibly , the more significant role of magnetic flux loops is to stir the icm in the center of cf clusters , by reconnection and motion caused by magnetic tension ( zoabi et al . 1996 ; zsr ) . this stirring , where some portions of loops falls fast and others buoyant outward , can lead to more reconnection between magnetic flux loops ( zoabi et al . 1996 ) , enhancing the efficiency of non - radiative cooling via local heat conduction as discussed by sbs . if the general scenario discussed here and/or in sbs holds , then a cf model with the following ingredients emerges . the evidences for cf in many clusters were summarized by soker et al . ( 2001 ) prior to the new results of the _ chandra _ and xmm - newton x - ray telescopes . the new results show that the cooling rate to low temperatures is much lower than the expected values in the old versions of the cf model , but they do show that some cooling occurs ( e.g. , a few clusters in kaastra et al . this is compatible with the moderate cf model . on the other hand , in most cf clusters the new x - ray observations did nt reveal any violent heating events , e.g. , major mergers or strong shocks , and the icm structure is not consistent with global heat conduction from the outer hot massive clusters regions . the x - ray observations do show heating , e.g. , bubbles and sound waves ( e.g. , fabian et al . 2003 ) , but these ca nt compensate for the cooling in many cases ( e.g. , kawano et al . observations using the high spatial resolution of _ chandra _ show most activity to occur in the very inner region , on a scale much smaller than the cooling radius ( where radiative cooling equals the cluster age ) , compatible with much lower mass cooling rate . all these suggest that the new x - ray observations actually strengthen the cf model , although a moderate one , rather than `` kill '' it . soker & david ( 2003 ) show how a non - steady state situation can account for the new observations . such a non - steady state evolution is expected in intermittent - heating models , with long enough time intervals . the gas is heated up , and then cools . most of the heated gas does not cool to low temperatures before the next major heating event . soker & david claim that no fine tuning is required . more accurate analysis may show that this non - steady state evolution can hold for heating events with shorter intervals as well . although not strictly a necessary ingredient , non - radiative cooling of the x - ray emitting gas after it cooled already by a factor of @xmath103 from its initial temperature , makes the agreement of the moderate cf with observations much better ( sbs ; this paper ) . the initial radiative cooling is required , most likely , to enhance by a factor of @xmath104 the magnetic pressure relative to the thermal pressure ( sbs ) . the magnetic fields play a crucial role in the scenario discussed here , and in the more efficient non - radiative cooling scenario studied by sbs . the outflow ( buoyant bubbles and magnetic flux loops ) and inflow , the strong magnetic fields , the possible importance of local heat conduction , radiative cooling in optically thin plasma , the presence of x - ray emitting gas together with optically emitting gas , all these and more , make the icm in cf clusters similar in some aspects to the gas above the solar ( and similar stars ) photosphere , i.e. , photosphere to corona regions . the similarity with respect to some processes relevant to magnetic field reconnection and optical line emission was mentioned by jafelice & friaca ( 1996 ) and godon et al . sparks & donahue ( 2003 ) note the morphological similarity between filaments in m 87 and the solar atmosphere . godon et al . ( 1998 ) discuss magnetic field amplification in cf clusters and the possibility of `` cluster flares '' , and use hints from solar flares to project on the expected nature of violent magnetic reconnection in cf clusters . some aspects of magnetic flux loop dynamics where studied by zoabi et al . ( 1996 ) and zsr . kaastra et al . ( 2003 ) extend the comparison of icm in cf cluster with stellar coronae to explain the x - ray spectra they observed in cf clusters . binney , j. 2003 , in `` the riddle of cooling flows in galaxies and clusters of galaxies '' , eds . reiprich , t. h. , kempner , j. c. , & soker , n. , charlottesville , va , usa , may 31-june 4 , 2003 ( astro - ph/0310222 ) fabian , a. c. , allen , s. w. , crawford , c. s. , johnstone , r. m. , morris , r. g. , sanders , j. s. , & schmidt , r. w. 2002 , mnras , 332 , l50 fabian , a. c. , mushotzky , r. f. , nulsen , p. e. j. , & peterson , j. r. 2001 , mnras , 321 , l20 fabian , a. c. , sanders , j. s. , allen , s. w. , crawford , c. s. , iwasawa , k. , johnstone , r. m. , schmidt , r. w. , & taylor , g. b. 2003 , mnras , 344 , l43 kaastra , j. s. et al . 2003 , a&a , in press ( astro - ph/ 0309763 ) . kahn , s. m. , peterson , j. r. , paerels , f. b. s. , xu , h. , kaastra , j. s. , ferrigno , c. , tamura , t. , bleeker , j. a. m. , & jernigan , j. g. 2003 , ( astro - ph/0210665 ) osullivan , e. , & vrtilek , j. m. 2003 , in `` the riddle of cooling flows in galaxies and clusters of galaxies '' , eds . reiprich , t. h. , kempner , j. c. , & soker , n. , charlottesville , va , usa , may 31-june 4 , 2003 ( astro - ph/0310099 ) peterson , j. r. , kahn , s. m. , paerels , f. b. s. , kaastra , j. s. , tamura , t. , bleeker , j. a. m. , ferrigno , c. , & jernigan , j. g. 2003a , apj , 590 , 207 peterson , j. r. , kahn , s. m. , paerels , f. b. s. , kaastra , j. s. , tamura , t. , bleeker , j. a. m. , ferrigno , c. , & jernigan , j. g. 2003b , in `` the riddle of cooling flows in galaxies and clusters of galaxies '' , eds . reiprich , t. h. , kempner , j. c. , & soker , n. , charlottesville , va , usa , may 31-june 4 , 2003 , soker , n. , blanton , e. l. , & sarazin , c. l. 2003 , in `` the riddle of cooling flows in galaxies and clusters of galaxies '' , eds . reiprich , t. h. , kempner , j. c. , & soker , n. , charlottesville , va , usa , may 31-june 4 , 2003 ( astro - ph/0309633 ) ( sbs ) sparks , w. b. , & donahue , m. 2003 in `` the riddle of cooling flows in galaxies and clusters of galaxies '' , eds . reiprich , t. h. , kempner , j. c. , & soker , n. , charlottesville , va , usa , may 31-june 4 , 2003	i study non - radiative cooling of x - ray emitting gas via heat conduction along magnetic field lines inside magnetic flux loops in cooling flow clusters of galaxies . i find that such heat conduction can reduce the fraction of energy radiated in the x - ray band by a factor of @xmath0 . this non - radiative cooling joins two other proposed non - radiative cooling processes , which can be more efficient . these are mixing of cold and hot gas , and heat conduction initiated by magnetic fields reconnection between hot and cold gas . these processes when incorporated into the moderate cooling flow model lead to a general cooling flow model with the following ingredients . ( 1 ) cooling flow does occur , but with a mass cooling rate @xmath1 times lower than in old ( @xmath1 years ago ) versions of the cooling flow model . namely , heating occurs such that the effective age of the cooling flow is much below the cluster age , but the heating ca nt prevent cooling altogether . ( 2 ) the cooling flow region is in a non - steady state evolution . ( 3 ) non - radiative cooling of x - ray emitting gas can bring the model to a much better agreement with observations . ( 4 ) the general behavior of the cooling flow gas , and in particular the role played by magnetic fields , make the intracluster medium in cooling flow clusters similar in some aspects to the active solar corona .
originally , the casimir effect was predicted as a feature of the electromagnetic field between two neutral ideally conducting plates and consisted in the appearance of an attractive force between the plates . the force is due to the change of the zero - point energy of the field in the confined space @xcite . in this special case , however , the casimir force can also be viewed as the long - range van der waals force @xcite . it becomes appreciable in the submicron and rapidly increases in the nanometer range . as such , it may strongly affect processing in nanotechnology as well as functioning of micro- and nanomachines and devices @xcite . clearly , these new developments pose the problem of realistic calculations of the casimir force on objects in complex enviroments . in contrast to the highly idealized system considered by casimir @xcite , lifshitz @xcite calculated the force between two thick ( semi - infinite ) dielectric slabs by taking into account the dispersion and absorption in the dielectrics as well as the temperature effects . in this respect , his theory is far more realistic and , as the effects of finite conductivity and dissipation in the metal can be observed in the recent high - precision experiments @xcite , his result for the force at zero temperature is standardly used when analyzing the casimir force in the planar geometry @xcite . the lifshitz approach is based on the calculation of the electromagnetic field due to the randomly fluctuating currents in the dielectric slabs and on the subsequent calculation of the maxwell stress tensor in the region inbetween . owing to its complexity , however , it has never been extended to the calculation of the force between multilayered stacks although the generalization of the final result to this configuration is fairly obvious . the casimir effect in multilayered systems is usually considered using either the surface mode summation method @xcite ( see also refs . @xcite ) to calculate the change in the electromagnetic field zero - point energy due to the presence of the dielectric stacks , or the stress tensor method @xcite to calculate directly the vacuum field pressure on the stacks . strictly speaking , the mode summation method applies only to purely dispersive ( lossless ) systems as only in this case the mode frequencies are real . however , when expressed as an integral over the imaginary frequency , the final result for the casimir energy ( force ) turns out to be applicable to absorbing systems as well . an indication that this must be so is the fact that the dielectric function is always real on the imaginary axis irrespective of whether the system is absorbing or not @xcite . thus , while in their calculation of the force in a multilayer zhou and spruch @xcite assumed a purely dispersive system , klimchetskaya _ et al_. @xcite recently considered a similiar but absorbing system . on the other hand , being a local approach , the stress tensor method does not necessarily imply a lossless system . since the stress tensor can not be defined macroscopically for an absorbing medium @xcite , the only necessary assumption is actually that the region where the vacuum - field pressure is calculated is nonabsorbing , whereas the other parts of the system may generally be dissipative . despite this fact , numerous papers in the past used the stress tensor method to calculate the casimir force assuming , at most , a dispersive but nonabsorbing system . one of the reasons for that is certainly lack of knowledge on the proper form and properties of macroscopic field operators aproppriate for an absorbing system at that time . the first calculation of the casimir force between two absorbing slabs is due to kupiszewska @xcite who modeled dielectric atoms as a collection of harmonic oscillators coupled to a heat bath that absorbs energy . only the modes propagating normally to the slabs were considered , so that this approach was effectively one - dimensional ( 1d ) . describing the reservoir through a damping constant and the langevin force and solving for the field operators , kupiszewska obtained for the force between the slabs the same expression in terms of their reflection coefficients as that obtained previously for an inert @xcite or a lossless @xcite 1d system , except that this time the dielectric function of the slabs was complex . recently , this result was rederived using a green function method for quantizing the macroscopic field in ( 1d ) absorbing systems in conjuction with a scattering matrix approach @xcite and was also extended to two identical absorbing superlattices @xcite . very recently , esquivel - sirvent _ et al_. demonstrated an alternative green function approach that makes the quantization of the field within the slabs unnecessary and calculated the casimir force in an asymmetric configuration @xcite which was earlier considered only in the lossless case @xcite . owing to their complex structure , an explicit calculation of the field operators as in refs . @xcite is highly impractical in the general case of a three - dimensional ( 3d ) dissipative inhomogeneous system . however , as pointed recently by matloob @xcite ( see also ref.@xcite ) , using the fluctuation - dissipation theorem and the linear response theory , the field correlation functions needed to calculate the stress tensor can be expressed in terms of the ( classical ) green function for the system . in this way , only the knowledge of the green function is , therefore , actually needed to calculate the casimir force . using this method , matloob and falinejad recently considered the casimir force between two identical absorbing dielectric slabs @xcite . very recently et al_. @xcite generalized their green function method @xcite to three dimensions and calculated the casimir force between two arbitrary slabs . expressing the reflection coefficients of the slabs through the generalized surface impedances , these authors argued that their formal result could be applied to rather general but not chiral media , also including nonlocal inhomogeneous dissipative slabs . in refs . @xcite the space between the slabs was assumed empty . in this work we calculate the casimir force in a lossless dispersive layer of an otherwise absorbing multilayer by employing the macroscopic field operators as emerge from a recently developed scheme for quantizing the electromagnetic field in inhomogeneous dissipative 3d - systems @xcite and using a convenient green function for a multilayer @xcite . in this way , we obtain a general result for the casimir force in stratified local media . in addition , using the properties of the generalized fresnel coefficients , we derive a relationship between the casimir force and energy in two different layers and demonstrate the applicability of the theory to more complex planar systems by calculating the casimir force on a diectric slab in a realistic planar cavity . consider a multilayered system described by the dielectric function @xmath0 defined in a stepwise fashion , as depicted in fig . 1 . the casimir force in a layer corresponds to the net vacuum - field pressure in the multilayer with respect to the pressure in the infinite layer ( medium ) . accordingly , the force @xmath1 on a stack of layers that separates a @xmath2th and an @xmath3th layer is given by @xcite @xmath4 where @xmath5 is the @xmath6 component of the regularized stress tensor in the @xmath2th layer @xmath7 with @xmath8 and @xmath9 being the corresponding maxwell stress tensors in the multilayer and in the infinite medium ( @xmath2 ) , respectively . ( [ flj ] ) , @xmath10 is the area of the stack and the @xmath11 ( @xmath12 ) sign applies if @xmath13 ( @xmath14 ) . since the regularized stress tensor vanishes in the outmost layers , we have for @xmath15 and @xmath16 : @xmath17 so that @xmath5 coincides with the force per unit area acting on the left ( right ) stack of layers bounding the layer ( @xmath2 ) . replacing field variables in the classical maxwell stress tensor @xcite by the corresponding heisenberg operators and taking its average @xcite , @xmath18 in a lossless layer @xmath19 [ @xmath20 is given by @xmath21 where we have suppressed the argument @xmath22 of the field operators and the brackets denote the expectation value in the vacuum state of the field . in order to calculate the correlation functions that appear in eq . ( [ t ] ) , we use the properties of the macroscopic field operators appropriate for absorbing systems @xcite . these operators are decomposed into their `` annihilation '' and `` creation '' components according to @xmath23 and , with the constitutive relations @xmath24 @xmath25 obey the standard macroscopic maxwell equations . here @xmath26 and @xmath27 are the noise polarization operators related to the dissipation in the system and obeying the commutation rules ( in the dyadic form ) : @xmath28 = \frac{\hbar\varepsilon''({\bf r},\omega)}{4\pi^2 } \tensor{\bf i}\delta({\bf r}-{\bf r}')\delta(\omega-\omega ' ) , \label{pp}\ ] ] where @xmath29 is the unit dyadic . therefore , any ( annihilation ) field operator is related to @xmath26 via the classical green function @xmath30 satisfying @xmath31 \tensor{\bf g}({\bf r},{\bf r'};\omega)=4\pi\tensor{\bf i } \delta({\bf r}-{\bf r ' } ) \label{gf}\ ] ] according to @xmath32 as a consequence , all field correlation functions can be expressed through the green function in accordance with the fluctuation - dissipation theorem @xcite . in particular , for the electric - field correlation function we have @xcite @xmath33 and the magnetic - field correlation function is easily obtained from this expression using @xmath34 . applying the above results to the @xmath2th layer and taking into account that @xmath35 is real and that @xmath36 in this region , we find for the relevant correlation functions in eq . ( [ t ] ) : [ cf ] @xmath37 @xmath38 where @xmath39 is the wave vector in the layer , @xmath40 is the green function element for @xmath41 and @xmath42 both in the layer @xmath19 , and @xmath43 is the corresponding green function element for the magnetic field . with the above equations inserted in eq . ( [ t ] ) , the stress tensor @xmath18 is expressed entirely in terms of the green function and its derivatives analogously to eq . ( [ tr ] ) below . similarly , through eqs . ( [ t ] ) and ( [ cf ] ) applied to the infinite medium ( @xmath2 ) , the stress tensor @xmath44 is given by the same expression with the infinite - medium green function @xmath45 . therefore , the regularized stress tensor @xmath5 is expressed as @xmath46 \nonumber\\ & & \left.+g^{b,\rm sc}_{j , zz}({\bf r},{\bf r};\omega)- g^{b,\rm sc}_{j,\parallel}({\bf r},{\bf r};\omega)\right\ } , \label{tr}\end{aligned}\ ] ] where @xmath47 is the green function for the scattered field in the @xmath2th layer and @xmath48 is its parallel trace . a convenient form of @xmath49 for a general multilayer is derived in ref . @xcite . in the appendix we quote this green function and calculate the expression in the curly brackets of eq . ( [ tr ] ) . inserting eq . ( [ cb ] ) , we find @xmath50 where @xmath51 , @xmath52 and @xmath53 are the reflection coefficients of the right and left stack of layers bounding the @xmath2th layer . the second line in eq . ( [ trf ] ) has been obtained by converting the integral over the real @xmath54-axis to one along the imaginary @xmath54-axis in the usual way , letting @xmath55 , @xmath56 and noting that the expression in the brackets is real on the imaginary axis . we see that the regularized stress tensor is uniform across the layer . although expected on invariance grounds , this is not a trivial result and , as is clear from the derivation in the appendix , it is due to cancellation of the @xmath57-dependent terms in the electric and magnetic contributions to @xmath5 irrespective of the dielectric properties of the surrounding stacks . knowing the force , the casimir energy @xmath58 in the layer can be calculated using @xmath59 with the condition that @xmath60 for @xmath61 . from eqs . ( [ trf ] ) and ( [ f ] ) , we find @xmath62 this equation , as well as that for the force [ combined eqs . ( [ f ] ) and ( [ trf ] ) ] , agrees in form with the corresponding result of zhou and spruch @xcite derived using the ( surface ) mode summation method and starting from a simple model of a purely dispersive multilayer . however , in this work the contributions of all ( propagating and evanescent ) modes are naturally taken into account on an equal footing through the green function . furthermore , since the green function employed refers to a general absorbing multilayer , so do the obtained results except , of course , for the region where the casimir force is calculated . the casimir energy and force vary in a stepwise fashion across the multilayer and we end this section by pointing out a relationship that exists between their values in two different layers , say , layers ( @xmath2 ) and ( @xmath3 ) . indeed , assuming that @xmath13 , for example , and using recursion relations for the reflection coefficients given by eq . ( [ rijk ] ) , one may prove that the following relation exists between the @xmath63 functions for the layers @xcite : @xmath64 combining this with eq . ( [ ej ] ) , we find that the respective casimir energies are related according to @xmath65_{\omega = i\xi}. \label{elej}\end{aligned}\ ] ] a similar relation is obtained for the forces in two layers , but the resulting expression is not particularly illuminating unless @xmath66 . such a situation arises , for example , when a planar object is embedded in a planar cavity . in this case , we find @xmath67_{\omega = i\xi } , \label{fljd}\end{aligned}\ ] ] where @xmath68 ( @xmath69 ) is the perpendicular wave vector in both layers . most of the previously obtained results for the casimir force and energy in a specific planar configuration are recovered from the results derived in the preceding section simply by specifying the corresponding reflection coefficients and material parameters . thus , for example , the results for the three - layer ( @xmath70 ) configuration considered by lifshitz @xcite are obtained letting @xmath71 , @xmath72 and @xmath73 , where @xmath74 are single - interface reflection coefficients given by eq . ( [ sic ] ) , and the results for the five - layer ( @xmath75 ) configuration considered by zhou and spruch @xcite are obtained letting @xmath71 , @xmath76 and @xmath77 , where the three - layer reflection coefficients are obtained from recurrencies eq . ( [ rijk ] ) . similarly , the results for the system consisting of two identical slabs , recently considered by matloob and falinejad @xcite , are obtained letting @xmath78 and @xmath79 , where @xmath80 are the reflection coefficients of a symmetrically bounded slab [ see eq . ( [ rt ] ) below ] , etc . specially , the casimir force and energy in a ( dispersive ) planar cavity formed by two ideally reflecting ( conducting ) slabs are obtained with @xmath81 . we also note that these equations correctly reproduce the corresponding results which emerge from the 1d considerations . indeed , taking only the @xmath82 contribution in eq . ( [ trf ] ) , we find from the first line in that equation , for example , @xmath83 where @xmath84 [ eq . ( [ dj ] ) ] is the same for both polarizations . with a simple algebra , this equation can be rewritten as @xmath85 , \label{trf1d}\end{aligned}\ ] ] which is in accordance with the casimir force obtained by several authors for the respective systems they considered @xcite . owing to the recursion relations which the generalized fresnel coefficients satisfy , the obtained results can be applied to more complex systems with planar symmetry . as an application of the theory , we illustrate this by deriving the casimir force on a dielectric , or a metallic , slab [ dielectric function @xmath86 , thickness @xmath3 ] in a cavity [ dielectric function @xmath88 with realistic mirrors [ reflection coefficients @xmath89 and @xmath90 , as depicted in fig . the force on the slab @xmath91 in this configuration can be calculated from eq . ( [ fljd ] ) . the function @xmath92 [ eq . ( [ dj ] ) ] is straightforwardly obtained letting @xmath93 and @xmath94 and using eq . ( [ rijk ] ) to determine the reflection coefficients @xmath95 . we find ( the polarization index @xmath96 is omitted ) @xmath97 where @xmath98 and @xmath99 are fresnel coefficients for the slab . this gives @xmath100^q_{\omega = i\xi},\nonumber\\ & n&=1-r(r_1e^{2i\beta d_1}+r_2e^{2i\beta d_2})\nonumber\\ & & + d(r^2-t^2)r_1r_2e^{2i\beta ( d_1+d_2 ) } , \label{f12}\end{aligned}\ ] ] where the expression in the brackets is to be calculated for @xmath96-polarization . using eqs . ( [ rrel ] ) and ( [ sic ] ) , @xmath101 and @xmath102 can be further expressed entirely in terms of the reflection coefficient for the cavity - slab interface @xmath103 [ where @xmath104 and @xmath105 as @xmath106 note that for a perfectly conducting ( @xmath107 ) slab , we have @xmath108 and @xmath109 . the force @xmath110 , as given by eq . ( [ f12 ] ) , may be positive or negative , depending on the dielectric properties of the slab and cavity mirrors as well as on the position of the slab . one may easily verify that this equation gives the correct result for the force on a perfectly conducting plate in an empty cavity with ideally reflecting walls . indeed , since in this case @xmath111 , n factorizes and eq . ( [ f12 ] ) splits to ( @xmath112 ) @xmath113 where @xmath114 ^ 3 } { e^{2\sqrt{\varepsilon(i\xi)}\xi d / c}-1 } \label{f(d)}\end{aligned}\ ] ] is the force on the left mirror of a dispersive ideal cavity [ cf . ( [ f ] ) and ( [ trf ] ) , with the index @xmath2 dropped ] . the second line here is obtained upon a partial integration over @xmath115 and upon calculating the @xmath115-derivative of the integral over @xmath116 ( see ref . @xcite ) . for the empty cavity [ @xmath117 , the integrals in eq . ( [ f(d ) ] ) become elementary giving the well - known result @xmath118 according to which the plate is attracted to the closer cavity mirror . for a partially transmitting plate , the vacuum - field fluctuations in regions ( 1 ) and ( 2 ) of the cavity are no longer independent of each other and considerable deviations from the above simple result may occur especially for realistic cavity mirrors . clearly , in this case , in order to explore the combined effect of the nearby walls on a planar ( nano)object , one must analyze eq . ( [ f12 ] ) numerically . using the properties of the macroscopic field operators appropriate for dissipative systems and a convenient green function for a multilayer , in this work we have obtained general results for the casimir force and energy applicable to local layered absorbing systems . we have also established a relationship between the casimir force ( and energy ) in two different layers and , as an application of the theory , calculated the casimir force on a dielectric slab in a realistic planar cavity . denoting the ( conserved ) wave vector parallel to the system surfaces by @xmath119 , we write the wave vector of an rightward ( leftward ) propagating wave in an @xmath3th layer as @xmath120 , where @xmath121 with this notation , the green function dyadic for the scattered field in the @xmath2th layer reads @xcite @xmath122\right . \nonumber\\&&\left.+r^q_{j+}e^{i\beta_j z_+}\hat{\bf e}^-_{qj}({\bf k } ) \left[\hat{\bf e}^-_{qj}(-{\bf k})e^{-i\beta_jz'_-}+ r^q_{j-}\hat{\bf e}^+_{qj}(-{\bf k})e^{i\beta_jz'_-}\right]\right\ } , \label{greensc } \nonumber\end{aligned}\ ] ] @xmath123 @xmath124 @xmath125 where @xmath126 are , respectively , the transmission and reflection coefficient of the upper ( lower ) stack of layers bounding the layer @xmath19 . clearly , for the outmost layers , @xmath127 , we have @xmath128 and @xmath129 . also , one must let @xmath130 since these quantities appear only formally . the remaining fresnel coefficients satisfy performing the derivations indicated in eq . ( [ gb ] ) and using @xmath137 we find that @xmath138 is given by eq . ( a2 ) multiplied by @xmath139 and with @xmath140 . noting that the equal - point green function dyadics consist only of diagonal elements , we easily find @xmath141\right.\nonumber\\ & & + \hat{\bf n}\hat{\bf n}\frac{\tilde{k}_j^2}{d_{sj } } \left[2r^s_{j-}r^s_{j+}e^{2i\beta_j d_j}+r^s_{j-}e^{2i\beta_j z_- } + r^s_{j+}e^{2i\beta_j z_+}\right]\nonumber\\ & & \left.+\hat{\bf z}\hat{\bf z}\frac{k^2}{d_{pj } } \left[2r^p_{j-}r^p_{j+}e^{2i\beta_j d_j}+r^p_{j-}e^{2i\beta_j z_- } + r^p_{j+}e^{2i\beta_j z_+}\right]\right\},\nonumber\end{aligned}\ ] ] and @xmath142 is given by this equation multiplied by @xmath143 and with @xmath144 . the traces @xmath145 and @xmath146 can be easily recognized from these equations and one has , for example , @xmath147= \frac{i}{2\pi}\int\frac{d^2{\bf k}}{\beta_j } \times\\ & & \left\{\frac{\beta_j^2}{d_{pj } } \left[2r^p_{j-}r^p_{j+}e^{2i\beta_j d_j}-r^p_{j-}e^{2i\beta z_- } -r^p_{j+}e^{2i\beta z_+}\right]\right.\nonumber\\ & & + \frac{\tilde{k}_j^2}{d_{sj } } \left[2r^s_{j-}r^s_{j+}e^{2i\beta_j d_j}+r^s_{j-}e^{2i\beta_j z_- } + r^s_{j+}e^{2i\beta_j z_+}\right]\nonumber\\ & & \left.-\frac{k^2}{d_{pj } } \left[2r^p_{j-}r^p_{j+}e^{2i\beta_j d_j}+r^p_{j-}e^{2i\beta_j z_- } + r^p_{j+}e^{2i\beta_j z_+}\right]\right\},\nonumber\end{aligned}\ ] ] while @xmath148 is given by this equation with @xmath144 . adding these two quantities , one finds that the curly bracket in eq . ( [ tr ] ) is equal to @xmath149 . \label{cb}\end{aligned}\ ] ] for new developments in the casimir effect and an extensive list of references see m. bordag , u. mohideen , g. l. klimchitskaya , and v. m. mostepanenko , phys . rep . * 353 * , 1 ( 2002 ) . s. k. lamoreaux , phys . lett . * 78 * , 5 ( 1997 ) ; _ ibid _ * 81 * , 5475(e ) ( 1998 ) .	the casimir effect in a dispersive and absorbing multilayered system is considered adopting the ( net ) vacuum - field pressure point of view to the casimir force . using the properties of the macroscopic field operators appropriate for absorbing systems and a convenient compact form of the green function for a multilayer , a straightforward and transparent derivation of the casimir force in a lossless layer of an otherwise absorbing multilayer is presented . the resulting expression in terms of the reflection coefficients of the surrounding stacks of layers is of the same form as that obtained by zhou and spruch for a purely dispersive multilayer using the ( surface ) mode summation method [ phys . rev . a * 52 * , 297 ( 1995 ) ] . owing to the recursion relations which the generalized fresnel coefficients satisfy , this result can be applied to more complex systems with planar symmetry . this is illustrated by calculating the casimir force on a dielectric ( metallic ) slab in a planar cavity with realistic mirrors . also , a relationship between the casimir force and energy in two different layers is established .
albeit rare , massive stars are the main contributors to the input of ionizing photons , energy , and momentum into the interstellar medium , and are responsible for a significant fraction of the chemical enrichment of their host galaxy . massive stars evolve on relatively short timescales ( a few @xmath14 years ) , and their evolution is strongly influenced by their strong mass loss and rotation ( e.g , * ? ? ? tremendous advancement in the understanding of the evolution of massive stars has been achieved in recent decades @xcite . in particular , we now have insights into the effects of rotation on the evolution of massive stars @xcite . the models predict the existence of a short - lived , transitional stage , usually referred to as the luminous blue variable ( lbv ) phase @xcite , during which the star has a high mass - loss rate ( @xmath15 @xmath16 ) . in the current picture of stellar evolution , lbvs are rapidly evolving massive stars in the transitory phase from being an o - type star burning hydrogen in its core to a hydrogen - poor wolf - rayet , helium - core burning star @xcite . the lbv phase is likely an unavoidable evolutionary stage which massive stars with initial mass @xmath17 will experience during their lifetime before becoming wrs and exploding as a supernova @xcite . whether lbv progenitors have an upper limit for @xmath18 , as suggested by the latest evolutionary models @xcite , still remains to be seen . several lbvs , such as eta car @xcite , ag car @xcite , and the pistol star @xcite , among others ( see @xcite ) , are known to have bolometric luminosities in excess of @xmath19 , which suggests a progenitor with @xmath20 . following recent determinations that the mass - loss rates of o - type stars are a factor of 320 lower than those adopted by the evolutionary models cited above @xcite , and the detection of massive nebulae around lbvs , there is growing evidence supporting the idea that the evolution of massive stars is even more dominated by strong mass loss during the lbv phase than previously thought @xcite . surprisingly , some recent work suggests that some core - collapse sne have lbv progenitors @xcite , which dramatically enhances the cosmological importance of lbvs and poses a great challenge to the current paradigm of massive star evolution . some of these type - ii sne occurred just a few years after the progenitor suffered a giant mass ejection la eta carinae in the 1840 s @xcite , raising the following provocative hypothesis as to the near - term fate of eta car itself and other lbvs lbvs might fail to lose additional significant amounts of mass and , thus , never become a wolf - rayet star @xcite . only a dozen or so lbvs are known in the galaxy @xcite , and determining their fundamental parameters and evolutionary status is a key to understanding unstable massive stars and population iii objects in the high - redshift universe . ag carinae ( = hd 94910 , @xmath21=10h56m11.6s , @xmath22=-60d27min12.8s ) is one of the brightest and most famous lbv stars . @xcite classified ag car as a p cygni - type " star almost a century ago , while its photometric variability was discovered by @xcite . its high luminosity was first reported by @xcite , who also noticed the presence of a bipolar nebula around the central star . after these pioneering works , intense observing campaigns in the past decades have provided a wealth of observational data on ag car , showing that the star has photometric , spectroscopic , and polarimetric variability on timescales from days to decades ( @xcite ; see references in @xcite ) . the strongest variability occurs cyclically on timescales of decades and is known as the s - dor cycle , during which the visual magnitude of the star decreases to @xmath23 mag @xcite . the advent of non - lte , full line - blanketed radiative transfer codes , such as cmfgen @xcite , has allowed us to obtain precise fundamental parameters for ag car during the visual minimum epochs of the s - dor cycle and to gain insights into the nature of the star ( @xcite , hereafter paper i ) . following the determination of strikingly different stellar and wind parameters at different minima ( ) , we would like to address the following questions in the present paper . is the bistability mechanism present in ag car ? what was the initial mass , and what is the current mass of ag car ? is ag car close to the eddington limit ? if so , is it during visual minimum or on the way to visual maximum ? how does rotation modify such a scenario ? what is the present evolutionary status of ag car ? this paper is organized as follows . in sect . [ obsmod ] we briefly describe cmfgen , the radiative transfer code employed in the analysis of the observed spectra of ag car . a detailed comparison between the minimum phases of the s - dor cycle of ag car under the light of the bistability mechanism are considered in sect . [ compbi ] . in sect . [ rot ] we present evidence that ag car was rotating fast during the visual minimum of 19851990 . the evolutionary status , initial mass , and current mass of ag car are considered in sect . [ evol ] . an analysis of how the fundamental parameters of ag car change from visual minimum toward maximum of the s - dor cycle is provided in sect . [ modus ] . the conclusions of this paper are summarized in sect . [ conclusions ] . the photometric and spectroscopic observations of ag car during visual minimum were extensively described in and references therein , to which we refer the reader for further details . the detailed spectroscopic analysis of ag car during visual minimum was done using the radiative transfer code cmfgen @xcite and was presented in . cmfgen has been successfully used to study other lbvs and related objects @xcite . since the code has been extensively discussed in the aforementioned references and in , we concisely describe its main characteristics below . cmfgen assumes a spherical - symmetric , steady - state outflow and computes continuum and line formation in the non - lte regime . each model is defined by the hydrostatic stellar radius r@xmath24 , bolometric luminosity l@xmath24 , mass - loss rate @xmath25 , wind terminal velocity @xmath26 , stellar mass @xmath27 , and abundances z@xmath28 of the included species . the radius of the hydrostatic core is defined as @xmath29 , in order to avoid any effects due to the strong wind in the determination of @xmath30 . the distance to ag car can not be constrained from the spectroscopic analysis using cmfgen . throughout this work , we assume that ag car is located at @xmath31 @xcite . we refer to for further discussion on the effects of the distance on the derived parameters . cmfgen does not solve the hydrodynamic equations to determine the wind structure , and therefore , a velocity law @xmath32 needs to be assumed a priori . in cmfgen , the velocity law is parameterized by a beta - type velocity law which is modified at depth to smoothly merge with a quasi - hydrostatic structure at the sonic point . the velocity structure below the sonic point is iterated to fulfill the wind momentum equation with a maximum error of 10% for @xmath33 . systematic errors in the density structure below the sonic point might be present due to the proximity of the star to the eddington limit and the choice of @xmath27 , rotation rate , and viewing angle . cmfgen allows for the presence of clumping within the wind . this is accomplished by assuming a homogeneous wind at distances close to the star , and that clumps start to be formed at a given velocity @xmath34 . the wind achieves a volume - filling factor @xmath35 at large distances , as follows : @xmath36\,\,.\ ] ] full line blanketing is included consistently in cmfgen through the concept of superlevels , which groups similar energy levels into one single superlevel to be accounted for in the statistical equilibrium equations . the final atomic model included h , he , c , n , o , na , mg , al , si , ti , cr , mn , fe , co , and ni . synthetic spectra were computed in the observer s frame using cmf_flux @xcite , including the effects due to rotation . after comparing the lightcurve and the spectroscopic evolution during both visual minimum phases , it can be noted that the star presents different characteristics in these epochs . an inspection of the ag car lightcurve shows that the visual minimum phases of 19851990 and 20002003 are significantly different , especially regarding their duration . while the star was in a visual minimum phase for about 5 years between 19851990 , the duration of the following visual minimum of 2000 - 2003 was much shorter , between 2 and 3 years . there are also significant qualitative differences in the spectrum obtained at different visual minimum phases of ag car , which can be inferred by comparing spectral lines of , , and . figure [ comphotphases ] displays , for each visual minimum , the spectrum obtained during the epoch when the maximum value of @xmath0 was achieved , corresponding to 19851990 and 20002001 . figure [ comphotphases ] presents the spectra obtained in 1987 june , 1989 march , and 2001 april , centered in three key lines for which high - resolution spectra were available : h@xmath37 , 5876 , and 4686 . h@xmath37 emission , in both the line core and the electron - scattering wings , was much stronger in 2001 than in 19851990 , indicating that the wind was denser in 2001 . the p - cygni absorption component of h@xmath37 was much deeper in 2001 , while it was almost absent in 19851990 . this indicates that the @xmath38 energy level population of h was higher in 2001 , which was probably caused by the lower @xmath0 and , thus , increased amount of neutral h. we will come back to this point in sect . [ ionouter ] . the 4686 emission , on the other hand , was stronger in 19851990 than in 2001 , confirming that @xmath0 was higher in 19851990 than in 2001 . the line emission of 5876 was stronger in 2001 than in 19851990 , but this was due to the increased wind density and not because of a higher @xmath0 , since the latter would also imply increased 4686 emission , and the opposite is seen . the p - cygni absorption profiles of h@xmath37 and 5876 both indicate that the wind terminal velocity was lower in 2001 than in 19851990 . the detailed spectroscopic analysis using cmfgen models confirms the above scenario and allows us to make quantitative conclusions through the comparison of the physical parameters obtained in the epochs when the maximum stellar temperature was achieved in consecutive visual minimum phases . while @xmath39 was similar and @xmath30 was only 10% different , three other physical parameters were significantly different ; namely , @xmath40 , @xmath25 , and @xmath41 . notably , @xmath0 was approximately 5800 k higher in 19851990 than in 2001 , while @xmath26 was 23 times higher and @xmath42 roughly two times smaller in 19851990 than in 20002003 . based on our modeling , we suggest that the differences in the lightcurve and in the spectrum of ag car during consecutive visual minimum phases are caused by the strikingly different _ wind _ parameters due to a change in the underlying stellar parameters . _ this is contrary to the current ideas about the nature of the s dor cycles , where the visual minimum phase is representative of a quiescent phase of the lbv star . in this scenario , visual minimum phases should be equal to each other , which is not the case for ag car . _ thus , the questions to be answered are : why are the physical parameters different in consecutive visual minimum phases ? is there any intrinsic reason for this to occur ? the differences in @xmath0 , @xmath42 , and @xmath26 during the minimum phases of ag car are immediately reminiscent of the bistability mechanism seen in line - driven winds @xcite , which has already been suggested to be responsible for the variable mass - loss rate seen in ag car @xcite . before going further and analyzing whether our results agree with previous works , we shall briefly summarize the physics behind the bistability mechanism . there are two potential effects related to sudden changes in the stellar parameters , which are commonly referred to as bistability . first , according to the original proposition of @xcite when analyzing models for the lbv p cygni , the classical bistability mechanism is characterized by a change in the optical depth of the lyman continuum ( @xmath43 ) , which in turn changes the ionization stages of metals in the wind , causing a change in the amount of line driving , and thus , ultimately changing @xmath42 and @xmath41 . later , this concept was extended to a bistability _ jump _ by @xcite to explain the physical parameters of blue supergiants , suggesting that the winds of those stars have two different regimes of @xmath42 and @xmath41 , which switch abruptly from one to the other around 21,000 k. @xcite proposed that blue supergiants with @xmath44 could typically have @xmath42 2 times higher ( correspondingly , @xmath26 is 2 times lower ) than blue supergiants with @xmath45 . the physical explanation for this phenomenon is a change in the ionization structure of iron in the inner stellar wind @xcite , which changes the number of spectral lines able to absorb / scatter radiation and , thus , able to drive the wind . this in turn affects the value of @xmath42 and @xmath41@xcite . while the @xcite work referred to the wind of a single lbv star , @xcite and @xcite were analyzing _ different _ ob supergiants . both ideas were combined by @xcite , who suggested that the variable mass loss commonly inferred for lbvs is due to subtle changes in the line driving due to recombination / ionization of fe@xmath46 into fe@xmath47 and fe@xmath47 into fe@xmath48 . however , based on a detailed spectroscopic analysis with cmfgen of a large sample of ob stars , the idea of a jump " in @xmath42 and @xmath26 of ob supergiants around @xmath45 has been disputed by @xcite , who suggest a more continuous variation of @xmath42 and @xmath26 with temperature . second , there is also the wind instability , which causes sudden changes in the h@xmath37 line profiles , intensity of lines , and @xmath43 , for a relative small change in @xmath39 , @xmath27 , and/or @xmath0 . this effect is well known in lbvs and has been discussed in the literature for p cygni @xcite , hde 316285 @xcite , and eta carinae @xcite . interestingly , the wind instability and the classical bistability appear to occur at similar @xmath0 , causing feedback effects as discussed by @xcite . in order to check whether the bistability mechanism can explain the presence of different stellar and wind parameters in consecutive visual minimum phases of ag car , we first turn our attention to the underlying cause of the bistability mechanism , which is the change in @xmath43 . we analyzed the behavior of @xmath43 in selected models obtained for both minima , which are shown in fig . [ lymancontion ] . while we obtained @xmath49 throughout the wind during 19851990 ( i.e. optically thin ) , an increase by three orders of magnitude in @xmath43 was derived in models during 20002003 , making the lyman continuum completely optically thick . therefore , we first conclude that _ the different stellar parameters at the base of the wind of ag car in different visual minimum phases cause a huge change in @xmath43 , which in turn cause significant differences in @xmath42 and @xmath26 , implying a very different spectral morphology . _ in particular , we conclude that when ag car had @xmath50 ( i.e. , during 19851990 ) , the lyman continuum was optically thin , @xmath42 was @xmath51 times lower than in 2001 , and @xmath52 times higher than in 2001 , when the lyman continuum was optically thick and @xmath53 . our results confirm the findings from @xcite that ag car lies on different sides of the bistability regime at different epochs @xcite . we found that between 19851990 , ag car had @xmath54 k , which we interpret as being on the hot side of the bistability regime , while in 2001 ag car had @xmath55 k , which we interpret as the star crossing the bistability limit into the cool side . this result implies that the effective temperature for occurrence of the bistability in ag car is somewhere between 17,000 k and 22,800 k. based on our detailed cmfgen spectroscopic analysis , we suggest that this limit is closer to the earlier , around 17,000k 18,000 k , because in this case , small changes of 5% in @xmath56 , @xmath25 , and/or @xmath30 caused significant changes in @xmath43 , in the ionization structure of the wind , and in the emerging spectrum . we shall now discuss the iron ionization structure predicted by our cmfgen modeling in the inner wind of ag car , since the huge change in @xmath43 is expected to impact the iron ionization structure @xcite . the inclusion of several physical processes in the current generation of radiative transfer stellar wind codes such as cmfgen turned out to be crucial for the analysis of the ionization structure of the wind of early - type stars . in particular , the exact treatment of radiative transfer in the co - moving frame , the inclusion of full line blanketing , wind clumping , and consistent modeling of the wind and photosphere have all had a huge effect on the ionization structure of early - type stars ( e.g @xcite ) . we anticipate that the ionization structure of lbv stars will also be affected . the fe ionization structure of ag car during 19851990 and 20002003 obtained from the cmfgen models is presented in the bottom row of figure [ lymancontion ] . we noticed that the ionization structure of fe before the sonic point , which is around @xmath57 for all the above models and below which @xmath42 is set , is modified when @xmath43 changes . during 19851990 the dominant ionization stage was fe@xmath46 , while in the following visual minimum phase of 2001 there is a steady increase in fe@xmath47 as @xmath0 decreased . since fe@xmath47 has a higher opacity than fe@xmath46 , particularly in the region where the stellar flux is high , more momentum can be removed from the radiation field and transferred to the wind , causing a higher mass - loss rate . nevertheless , as seen in fig . [ lymancontion ] , fe@xmath46 is still the dominant ionization stage below the sonic point in the temperature regime found for ag car during visual minimum . @xcite proposed that the bistability mechanism is the cause of the variable behavior of the mass - loss of lbvs as a function of the effective temperature , predicting the presence of jumps in the @xmath42 vs. @xmath0 relationship at @xmath58 and @xmath59 due to the recombination of fe@xmath46 into fe@xmath47 and fe@xmath47 into fe@xmath48 , respectively . figure [ mdotvinfmin ] presents @xmath42 as a function of @xmath0 during the visual minimum epochs of ag car analyzed in paper i. is there clear evidence of a jump in @xmath42 as a function of @xmath0 as ag car crosses from one side of the bistability regime to the other ? from the point of view of the spectroscopic modeling , we can not give a definitive answer to this question for the following reasons . in general , we can see a progressive increase in @xmath42 when @xmath0 decreases in the range @xmath60 . close to the region where the lyman continuum becomes optically thick ( aka the bistability limit ) at @xmath61 , we can identify ag car in the two regimes predicted by @xcite . on the hot side , @xmath42 is lower than on the cool side , while @xmath26 is higher . however , whether the star switches abruptly from one regime to another or changes gradually can not be determined since ag car has never been observed at temperatures very close to the bistability limit . long - term spectroscopic monitoring has shown that @xmath0 of ag car decreases quickly from visual minimum to maximum ( on timescales of a few months , @xcite ; paper iii ) , and monitoring ag car in intervals of weeks would be required . even then , our ability to derive the stellar and wind parameters would likely be hampered by strong time - dependent effects , since the flow timescale during these epochs is also of the order of a few months . nevertheless , we can conclude from the short timescale that the stellar wind is highly unstable when close to the bistability limit , and the inner layers might be exposed to some sort of feedback from the unstable wind @xcite . the differences in our results from the predictions of @xcite are likely a combination three factors : a ) the inclusion of full line blanketing and wind clumping in our analysis , b ) the higher @xmath62 derived in our work compared to the value assumed by @xcite , and c ) due to the fact that the ratio @xmath63 is definitely not constant when @xmath0 changes , contrary to the assumption of @xcite . furthermore , @xmath42 and @xmath63 of ag car are greatly modified by rapid rotation @xcite , a phenomenon which was not included in the @xcite models . the influence of rotation will also be variable as a function of @xmath0 , since the ratio @xmath64 changes from visual minimum toward maximum ( @xcite ; see sect . [ rot ] ) . the behavior of @xmath26 as a function of @xmath0 is more irregular . first , there is a decrease in @xmath26 from @xmath65 until @xmath66 . then , a steep increase until @xmath67 is seen , and a shallow decrease again until @xmath68 . while the overall behavior of @xmath26 might be explained in terms of the bistability mechanism through changes in the line driving @xcite , there are a number of physical mechanisms which likely have a large impact on the behavior of @xmath26 on the cool side : the fast rotation derived in ag car , the presence of a latitude - dependent wind and departure from spherical symmetry , the accuracy of the velocity structure near the sonic point , and time - dependent effects . the presence of strong lines in the spectrum of ag car during the visual minimum of 20002003 ( which were absent during 19851990 ) provide solid evidence that the ionization structure of the _ outer _ wind of ag car was also different in those minima . figure [ agcionesttot ] presents the h and fe ionization structures in the whole wind of ag car during visual minimum phases as derived from the cmfgen models with the parameters from . it can be seen that in addition to the recombination of fe@xmath46 into fe@xmath47 in the inner wind during the 20002003 visual minimum , there is also recombination of h@xmath48 into h@xmath69 and of fe@xmath47 into fe@xmath48 in the outer wind of ag car . this explains the appearance of p - cygni absorptions during the latter visual minimum , which is only possible when a significant ( @xmath70 ) fraction of hydrogen is recombined . the same mechanism explains the absence of p - cygni absorption during 19851990 , since hydrogen was ionized throughout the whole wind . in addition , as first noted by @xcite and @xcite , the fe ionization structure in the outer wind is dominated by the following charge - exchange reaction with h , @xmath71 this reaction implies that hydrogen recombination is also responsible for the recombination of fe@xmath47 into fe@xmath48 , thus explaining why there are no strong fe ii lines in the 19851990 spectrum while they are relatively strong in data from 20002003 . the marked difference in the ionization structure of the outer wind during different visual minimum phases of ag car led us to investigate whether the recombination of hydrogen and fe@xmath47 into fe@xmath48 in the outer wind and the consequent appearance of a p - cyg absorption component are also due to the bistability mechanism ( i.e. change in the optical depth of the lyman continuum ) . we found that when @xmath43 increases , the relative population between the hydrogen levels 2 and 1 also increases , and so does the optical depth of the balmer lines , such as h@xmath37 and h@xmath72 , explaining the increase in the absorption of those lines . interestingly , variable p - cygni absorption has been observed in the lbv candidate he 3 - 519 @xcite , which might indicate that he 3 - 519 is also close to the bistability limit . further work is highly desirable in order to prove whether this speculation holds for lbvs in general . the analysis of broad absorptions due to 40884116 in the high - resolution ag car spectra obtained during the 20002003 visual minimum has shown direct evidence , for the first time , that some lbvs rotate at a high fraction of their critical rotational velocity @xcite . using a similar methodology , the bona - fide lbv hr carinae was also shown to possess a rotational velocity close to critical @xcite . since the lines were also present in ag car during the 19851990 visual minimum @xcite , we analyzed whether a similar fast rotation could be noticed during these epochs . exactly the same technique as described in @xcite was employed . although the spectra of ag car obtained in 1989 march have only moderate resolution around the lines ( r@xmath73100 @xmath74 ) , significant evidence of fast rotation can be inferred from the comparison between the observed spectrum and the 2-d models presented in fig . [ agcvrotr ] . the line profiles from models without rotation , convolved to the same spectral resolution as the observations , are narrower and have stronger central absorption than the observed line profiles of ag car from 1989 march . _ this is clear evidence that fast rotation is present . _ the best fit was achieved using @xmath75 , mainly using 4088 as a diagnostic ( fig . [ agcvrotr]a , ) since 4116 is severely blended with 4120 at the data resolution ( fig . [ agcvrotr]b ) . we estimate that @xmath76 likely had similar values during the whole 19851990 visual minimum , since there was little variability in the stellar parameters during these epochs . indeed , a tracing of a medium resolution photographic - plate spectrum of ag car obtained in 1985 @xcite shows broad 40884116 absorptions which are very similar to those predicted by our rotating models . figure [ agcvrotr]c presents the rotational velocity of ag car as a function of @xmath30 for each epoch . using a least - squares linear fit , we obtained @xmath77 , where the quoted error only includes the statistical errors . additional systematic errors , which are currently hard to estimate , might be present due to model assumptions . a flatter dependence of @xmath76 on @xmath30 might be obtained when including the deformation of the stellar surface from spherical symmetry , variation of the stellar and wind parameters as a function of latitude , and a more complex treatment of the density structure at and below the sonic point in the model . as a consequence , we can not exclude that @xmath78 . in addition , our result strongly suggests that magnetic fields are not important on the surface of ag car ; otherwise , significant deviations from the above relationship would be expected ( a. maeder , private communication ) . if the magnetic field on the hydrostatic surface of ag car played an important role , one would expect that the magnetic field would maintain the surface material corotating out to some distance . in this case , the surface rotational velocity would behave as that of a solid body , i.e. , increasing with @xmath30 . this is contradictory with our findings . the rotational period of ag car during 19851990 is @xmath79 days , assuming @xmath80 and @xmath81 . lbv stars have low - amplitude photometric variability ( the so - called micro - variability " ) in timescales of days to weeks @xcite . generally , the currently accepted mechanism for such behavior is non - radial pulsation @xcite . in the case of ag car , the micro - variability period during 19851990 was between 1114 days @xcite . the similarity between the values of @xmath82 and the micro - variability period is striking . we suggest that they might be related to each other and that the micro - variability is modulated by the rotational period of ag car , which would be the case if spots are present on the stellar photosphere , for instance . since the stellar photosphere of ag car is extended and located in the stellar wind , the presence of spots would suggest that structures propagate from the stellar surface into the wind , possibly causing effects on line profiles . if the photometric micro - variability is indeed modulated by rotation , we anticipate that the micro - variability period should be proportional to @xmath83 in the case of ag car , since @xmath76 is proportional to @xmath84 for ag car . if rotation is somehow related to the micro - variability of lbvs , in principle , their rotational velocities could be derived from the micro - variability period once the stellar radius @xmath30 is known . this analysis is beyond the scope of this paper and is the subject of an ongoing work . it is interesting to note that for other lbvs such as hr car and s dor , the variations in the micro - variability period between visual minimum and maximum is about 5 @xcite , which is consistent with a rotational modulation . in this section we compare the fundamental parameters obtained for ag car with the evolutionary tracks from @xcite ( hereafter mm03 ) . there are a couple of caveats which need to be considered : 1 ) the problematic definition of @xmath0 in lbvs and wrs because of the presence of an optically - thick wind and 2 ) the strong , long - term variability seen in ag car during the s - dor cycle . the first issue is a direct consequence of the high mass - loss rate of lbvs and wrs , which causes the presence of an extended photosphere . the hydrostatic base of the wind ( @xmath30 in our definition ) has a higher temperature ( @xmath85 ) than that of the photosphere ( @xmath0 , defined at @xmath86 ) . the relative decrease of @xmath0 compared to @xmath85 depends on how optically thick the wind is , which is a function of @xmath27 , @xmath42 , and wind velocity law ( e.g. , @xcite ) . additional complications come from the presence of wind clumping and time - dependent effects in lbvs , which are both not considered by evolutionary models . since the evolutionary models do not solve the radiative transfer in the extended stellar atmosphere as cmfgen does , they can not self - consistently compute @xmath0 during the lbv and wr phases . only a rough estimate of the effect of the optically - thick wind on @xmath0 can be calculated based on the stellar parameters and on the predicted mass - loss rate @xcite . obviously , that result will only be approximate and might not be accurate , as any changes in the assumed parameters ( such as @xmath42 ) will bias the comparison between @xmath0 determined from the cmfgen models and @xmath0 estimated by the evolutionary models . _ therefore , comparing the value of @xmath0 predicted by the evolutionary models with values obtained from the atmospheric radiative transfer codes might be misleading in the case of stars with dense winds such as lbvs and wrs . _ in this case , stars with a similar @xmath0 can have very different underlying stellar parameters at the base of the wind if they have different @xmath27 , @xmath42 , and/or @xmath26 . the second issue is related to the strong variability of ag car due to the s dor - type variability . which epoch of the s - dor cycle should be compared with the evolutionary tracks , since the models do not account for the s - dor type variability ? the answer to this question is not obvious . although the current ideas about the s - dor cycle consider the visual minimum phase the quiescent phase , followed by eruptions during maximum ( e.g. @xcite ) , there is much evidence which might not corroborate this point . first , we detected that @xmath6 in ag car is higher during visual minimum than during maximum ( sect . [ modus ] ) . therefore , if ag car is to settle in quiescence during visual minimum , it has to do so extremely close to the eddington limit . second , even if we assume that the star is at quiescence during visual minimum , ag car can have different stellar parameters in different minimum phases . which one should be compared with the evolutionary tracks ? third , some well - documented lbvs , such as s doradus @xcite and r127 @xcite , have been observed at maximum for extended time intervals . fourth , there is historical evidence that some lbvs , such as ag car , have presented the s - dor type instability for over a century @xcite , and the key point is how long the s - dor cycles are present . _ what if the s dor - type instability is present during the majority of the lbv phase for these stars ? _ in this case we would just be witnessing these unstable , massive stars evolving in real time , and they would never really be in quiescence during the lbv phase . a definitive answer to this point can only be provided when the physical mechanism behind the change of the hydrostatic radius of lbvs , which ultimately causes the s dor - type variability , is recognized and included in the stellar evolutionary models . keeping these caveats in mind , we present in figure [ agcevoltrack ] the location of ag car in the hr diagram during the visual minimum phases of 19851990 and 20002001 based on the parameters derived in , compared to those predicted by the evolutionary tracks for non - rotating and rotating stars from . the he abundance of ag car ( y=0.62 , ) is a strong constraint on the evolutionary stage of ag car , restricting the comparison to a very specific part of the evolutionary tracks . according to them , the bolometric luminosity of ag car is consistent with an initial mass of @xmath87 , independent of the initial rotational velocity . technically , only the non - rotating models in the range @xmath88 ( @xmath89 for rotating models ) have a bona - fide lbv phase ; in models with @xmath90 the star skips the lbv phase and goes directly to the wr phase . while both the rotating and non - rotating models can reproduce fairly well the effective temperature of ag car ( fig . [ agcevoltrack ] , lower panels ) , only the non - rotating model is consistent with the value of @xmath85 derived for ag car ( upper panels ) . the evolutionary models with rotation predict a much hotter ( @xmath91 ) wn star at the base of the wind , corresponding to a wn7 spectral type @xcite . even though the stellar evolution model with rotation can fairly reproduce the @xmath0 of ag car , it can not reproduce the value of @xmath85 that we inferred based on the spectroscopic analysis . therefore , based on current evolutionary models , we can rule out that the progenitor of ag car had a high initial rotational velocity . the non - rotating models simultaneously reproduce the values of @xmath85 , @xmath0 , and he / h ; the value of @xmath85 is consistent with a wn11 spectral type . however , the fast rotation of ag car is puzzling in the non - rotating scenario . one possibility is that the initial rotational velocity of ag car was significantly lower than 300 @xmath74 . another possibility is that additional processes need to be taken into account by the evolutionary models in order to fully explain the stellar properties during the lbv phase . for instance , it would be desirable to include giant outbursts of several @xmath92 in the evolutionary models to investigate the effects of this event on the evolution of an lbv . the influence of the stellar mass @xmath27 and effective gravity @xmath93 on the spectrum of ob supergiants is usually obtained from the broadening of h and he lines . however , the photosphere is extended when a dense stellar wind is present and , as the location of the photosphere approaches the sonic point , line emission from the wind fills out the original photospheric absorption profile . thus , in the case of lbvs and wr stars , the dense wind causes the photosphere to be located at velocities larger than the sonic point and strong emission lines appear in the spectrum . therefore , the influence of log @xmath93 through the analysis of hydrogen absorption lines is not applicable to ag car . to analyze the influence of log @xmath93 on the spectrum of ag car , we ran cmfgen models assuming different stellar masses in the range @xmath94 , keeping the other model parameters constant , such as a he mass fraction of 0.62 . the stellar mass affects the eddington parameter @xmath95 since @xmath96 where @xmath97 is the total wind opacity and the other symbols have their usual meaning . for epochs where the wind density was large enough that @xmath98 and the star was not so close to the eddington limit ( @xmath99 ) , such as during 2002 march 2002 july and 2003 january , we found little influence of log @xmath93 ( and hence @xmath27 ) on the emerging spectrum . the physical parameters derived from models with different log @xmath93 were virtually the same , and little insight on the value of @xmath27 could be obtained . this result is similar to what has been found for eta car @xcite and wr stars in general . however , for epochs when @xmath100 ( 19851990 and 20002001 ) , we found that the emerging spectrum changed significantly as a function of the adopted @xmath27 , and so did the parameters obtained from the spectroscopic analysis . unfortunately , @xmath27 could not be directly derived from the models , since a reasonable fit of the spectrum could still be obtained by adjusting other parameters such as @xmath0 , @xmath42 , and @xmath35 . we derived that the changes in @xmath27 affected mostly @xmath0 , which varied between 1960023800 k when @xmath27 varied from 40 to 150 @xmath101 , as shown in fig . [ gammamass ] . the value of @xmath0 is sensitive to the density and velocity structure below the sonic point , and we determined that @xmath0 is usually 10002000 k higher in models which adopted a fixed isothermal scale height to match a hydrostatic structure ( using a procedure described by @xcite ) . the dependence of @xmath0 on @xmath27 in this parameter regime can be understood as being due to the formation of a progressively more extended photosphere as the star approaches the eddington limit . when @xmath27 decreased ( while @xmath39 and @xmath0 were kept constant ) we found that higher ionization was achieved in the wind . therefore , a decrease in @xmath0 is needed to fit the observed spectrum again . the middle panel of fig . [ gammamass ] shows @xmath95 , obtained from our cmfgen modeling , as a function of @xmath27 , computed in two locations of the wind : at high optical depth ( @xmath102 and @xmath103 ) and at @xmath30 ( @xmath104 ) . according to our cmfgen models , ag car has to have @xmath105 to ensure @xmath106 . different relationships were obtained : while @xmath107 at @xmath108 , a much less steep variation was obtained at the base of the wind , at @xmath109 ( @xmath110 ) . the different steepness at different depths in the wind reflect the different photospheric extensions and opacities as a function of @xmath27 . the high rotational velocity derived for ag car during visual minimum ( @xcite ; sect . [ rot ] ) will reduce the maximum luminosity that can be reached before @xmath111 ( @xcite , hereafter mm00 ) and will require a larger value of @xmath27 in order for the star to be below the eddington parameter modified by rotation ( @xmath13 ) . here we show that , making a few reasonable assumptions , we are able to obtain a lower limit for @xmath13 and @xmath27 when taking rotation into account . according to , @xmath112}{4 \pi c g m \left(1-\frac{\omega}{2 \pi g \rho_\mathrm{m}}\right)}\,\,,\ ] ] where @xmath113 is the latitude angle and @xmath114 is a correction factor to the von zeipel gravity - darkening law to take into account the baroclinicity of the star . since the stellar parameters of ag car were obtained in through a one - dimensional radiative transfer analysis , and the high rotational velocity will cause deviations from spherical symmetry , there are a few caveats which need to be taken into account to obtain a reasonable value of @xmath13 . first , the mass - absorption coefficient and radiative flux may vary as a function of latitude in a fast - rotating star . unfortunately , given the 1-d nature of our radiative transfer analysis , we can not infer directly how these changes will impact the estimation of @xmath13 . to obtain a rough estimate of how close ag car is to the eddington limit modified by rotation ( the so - called @xmath6-limit ) , we ignore latitudinal variations of the mass - absorption coefficient and radiative flux in eq . [ gammaomega ] , which then reduces to @xmath115 to properly include realistic latitudinal variations of the opacity and radiative flux , one would need detailed hydrodynamical modeling coupled with radiative transfer calculations in 2-d , which is certainly well beyond the scope of this paper . at first glance , the assumption that the derived @xmath39 corresponds to the equatorial value of @xmath39 might imply that we are neglecting gravity - darkening , i.e , that the flux of a rotating star is larger at the polar direction than at the equator @xcite . however , the value of @xmath39 of a rotating star derived using a 1-d model also depends on the viewing angle @xcite . for instance , for a main - sequence o - star with @xmath116 rotating at 0.9 of its critical velocity for break - up , the actual value of @xmath39 is about 25% higher than the derived value of @xmath39 when observing equator - on and using a 1-d model ( warren & hillier , private communication ) . the presence of the wind will likely change this dependence . taking into account that ag car is seen nearly equator - on @xcite , @xmath39 obtained in is likely lower limit , and it is reasonable to assume that the flux at low latitudes is not much different from the one predicted by a spherical symmetric model with @xmath117 as in . equation [ gammaomega_1d ] can be rewritten as a function of @xmath76 as : @xmath118}\,\,,\ ] ] where @xmath119 is the equatorial radius at a given @xmath76 , @xmath120 and @xmath121 are respectively the equatorial and polar radius at the classical break - up , @xmath122 is the first critical velocity for rotation , and @xmath123 is the volume of the star relative to that of a sphere with @xmath121 , as in . assuming that the factor @xmath124 is close to 1 in ag car , and using the definition of @xmath125 , eq . [ gammaomega_1d_vrot ] reads as @xmath126}\,\,.\ ] ] the second caveat regards how @xmath30 , determined from the 1-d radiative transfer model , relates to @xmath121 . we argue that @xmath127 because 1 ) @xmath128 is at maximum 0.7 for ag car , meaning that the shape of the star does not significantly departs from spherical symmetry , and 2 ) the effects of fast rotation on the stellar structure were not included in cmfgen in the computation of the hydrostatic atmospheric structure below the sonic point . using this simplification , we can then replace @xmath121 by @xmath30 and , using @xmath95 and @xmath76 determined from the spectroscopic analysis and assuming @xmath129 , obtain @xmath13 from models with different @xmath27 using eq . [ gammaomega_1d_vrot_simpl ] . the results from this calculations are shown in the bottom panel of figure [ gammamass ] . since @xmath76 is considerable , the minimum mass of ag car will be larger than the value of @xmath130 derived in section [ massdep ] according to the @xmath95-limit . based on the 1-d cmfgen models we conclude that , ultimately , the fast rotation of ag car requires @xmath131 in order to satisfy the eddington limit modified by rotation . it would be desirable to confirm this result with appropriate full 2-d , line - blanketed , non - lte radiative transfer models when these become available . the first constraint to the current mass of ag car comes from evolutionary models . according to the non - rotating evolutionary models of , and assuming an initial mass of @xmath132 and the derived surface he abundance ( y=0.62 , ) , the current mass of ag car would be roughly @xmath133 . this value is extremely close to the minimum mass of ag car according to the modified eddington limit derived above , which would imply that ag car is _ at _ the @xmath134 limit . there are two important caveats of doing such a comparison . first , the evolutionary models of do not include giant eruptions , while ag car is known to have had a strong mass ejection @xmath135 years ago . second , the evolutionary models of include the mass - loss recipe during the main sequence ( ms ) and blue supergiant ( bsg ) phases from @xcite , while recent findings ( e.g. , @xcite ; see sect . [ intro ] ) seem to suggest a lower mass - loss rate during these epochs . if that is true , the evolutionary models would predict that ag car would reach the lbv phase with a lower @xmath27 . a second constraint to the current @xmath27 can be obtained by computing the mass budget throughout the evolution of ag car . if we take into account that ag car has a massive nebula ( @xmath136 ) composed of material from the star @xcite and an initial mass of @xmath132 , then the mass budget is tight : ag car would have @xmath137 assuming zero mass loss during the ms and bsg phases . a more realistic estimate of @xmath138 during @xmath139 years would lead to @xmath140 . if the mass - loss rate during the ms and bsg phases was even lower , we could make the case for ag car having @xmath141 . _ therefore , evolutionary considerations and the presence of a massive nebula suggest that ag car has @xmath142 and is , thus , dangerously close to the eddington limit modified by rotation_. obviously , a precise determination of the amount of nebular mass around ag car and detailed evolutionary models that include a lower mass - loss rate during the ms than the one currently used , and including the mass lost in brief giant eruptions , are highly desirable . the evolution of lbvs during their s - dor cycles is commonly assumed to be due to changes in the stellar radius at constant luminosity @xcite , which imply changes in the effective temperature . based on a detailed spectroscopic analysis of multi - epoch observations of ag car using cmfgen , we argue that this scenario needs some revisions in order to explain the temporal behavior of the physical parameters of ag car which are shown in fig . [ physparmin ] . we found that the stellar radius of ag car changes during the s - dor cycles , increasing from visual minimum to maximum , as has been determined by previous works ( e.g @xcite ) . however , unlike previous works , using cmfgen we can precisely derive both the hydrostatic radius of ag car ( @xmath143 , defined as @xmath144 ) and the photospheric radius ( @xmath145 , defined at @xmath86 ) for a given epoch , and thus , compare how they change as a function of time . we found that both @xmath30 _ and _ @xmath146 increase from visual minimum to maximum , ruling out that the s - dor type variability is caused only by an expanding pseudo - photosphere due to an increase in @xmath42 @xcite . the increase in @xmath42 is not high enough to explain the changes in @xmath146 ; definitely , our results show that a physical expansion of @xmath30 is needed to explain the s - dor type variability from visual minimum to maximum in the lightcurve , as first suggested by @xcite . the relatively high @xmath42 derived for ag car compared to b supergiants in the same temperature range , in combination with the proximity of the star to the eddington limit , causes the unavoidable formation of an extended pseudo - photosphere , which translates into different values for @xmath30 and @xmath146 . a major result from is that the bolometric luminosity of ag car decreases during the expansion of the hydrostatic radius , which is contrary to the aforementioned _ status quo _ explanation for the s - dor variability . the detection of a decrease in @xmath39 is not unique to ag car ; @xcite also detected that another famous , highly - variable lbv , s doradus , presents a similar behavior . we support the hypothesis by @xcite that the decrease in @xmath39 is due to the energy needed to expand the outer layers of the star . variations in @xmath39 were also detected in the lbv afgl 2298 @xcite . however , in this latter case , the variation of @xmath39 were apparently not related to changes in @xmath30 . another striking result is that the eddington parameter @xmath95 of ag car _ decreases _ as the star expands and its @xmath39 and @xmath0 decrease ( fig . [ physparmin ] ) . such a reduction is , at least in part , caused by the decrease in @xmath39 . for ag car , @xmath147 during the hottest epochs , and decreases to @xmath148 as the star evolves toward maximum . the eddington parameter modified by rotation @xmath6 is even more dramatically reduced from visual minimum to maximum ( fig . [ physparmin ] ) , since rotation will be relatively more important when the star has a higher @xmath0 ( see discussion in @xcite ) . while ag car is significantly close to the eddington limit modified by rotation ( @xmath149 during 19851990 and 2001 april ) during visual minimum , we found that @xmath150 as the star expands . therefore , if the proximity to the eddington limit reflects the instability of the outer layers of ag car , the star is more unstable during visual minimum than during maximum . the evolution of the wind parameters of ag car is regulated by the bistability mechanism ( sect . [ compbi ] ) . however , the behavior of the mass - loss rates and wind terminal velocities obtained in do not closely follow the predictions of @xcite . this is due to several reasons : our modeling determined that ag car is twice as massive as the value derived by @xcite , more luminous , and closer to the eddington limit . furthermore , ag car has a high rotational velocity which varies from visual minimum to maximum @xcite . since the models from @xcite do not include rotation , significant differences are anticipated . the detailed spectroscopic analysis of ag car using the radiative transfer code cmfgen has provided a wealth of information on the lbv phenomenon and on the s - dor type variability . below , we summarize the main conclusions of this paper . \1 . the high rotational velocity of ag car detected by @xcite during the visual minimum of 20012003 was also obtained during the previous visual minimum of 19851990 , amounting to @xmath151 . we found that the rotational velocity is proportional to @xmath152 , but including systematical errors which might arise due to the star being extremely close to the eddington limit , we can not rule out that @xmath153 . we noticed that the rotational period of ag car during the visual minimum of 19851990 ( @xmath154 days ) is remarkably similar to the cyclical photometric micro - variability of 1114 days observed during these visual minimum epochs @xcite . we suggest that the micro - variability period of ag car might be modulated by the stellar rotation . if rotation is also responsible for the micro - variability seen in other lbvs , in principle , their rotational velocities can be derived using this method . we found in that the consecutive visual minimum phases of 19851990 and 20002003 are different in duration , maximum magnitude achieved , maximum stellar temperature , mass - loss rate , and wind terminal velocity . we suggest that these differences arise due to different stellar parameters , which cause different mass - loss rates and wind terminal velocities . we explain this behavior in terms of the bistability mechanism of line - driven winds , which is ultimately regulated by the optical depth of the lyman continuum . from 19851990 , ag car was on the hot side of the bistability ( @xmath40=22800 k ) , while the star was on the cool side ( @xmath40=17000 k ) from 20002003 . therefore , ag car is the first star ever confirmed to be on different sides of the bistability at different epochs based on detailed spectroscopic analysis . we noticed that ag car has never been observed exactly _ at _ the boundary of the bistability limit , suggesting that significant instability is present when that happens . although we identify ag car in two regimes , the absence of observations close to the boundary of crossing from one side to the other of the bistability does not allow us to confirm the presence of a @xmath155 in the wind properties of ag car . an increase in @xmath42 as a function of @xmath0 was found as the star crosses the bistability limit , while @xmath26 decreases and presents a more complex behavior . the duration of the last two visual minimum phases of ag car and the variability inside the visual minimum are related to the maximum temperature achieved at each minimum and , therefore , on which side of the bistability the star was . it seems that visual minimum phases that have @xmath0 high enough to put the star on the hot side of the bistability limit have a longer duration , while visual minimum phases which have @xmath0 low enough to put the star on the cool side , are brief . this behavior needs to be verified by further spectroscopic and photometric monitoring of ag car during the next decades . we suggest that the proximity to the eddington limit modified by rotation ( @xmath6 ) plays a key role in explaining the behavior of ag car . the high value obtained for its luminosity ( @xmath5 ) , together with the high rotational velocity of ag car detected by @xcite and by this work , put the star extremely close to the eddington limit modified by rotation , @xmath6 @xcite . for an assumed mass of 70 @xmath92 , @xmath149 during 19851990 and 2001 , and decreases to @xmath156 as the star expands towards maximum . \7 . despite the extreme conditions found in ag car due to its high luminosity and fast rotation , the mass - loss rate remains relatively modest and does not wildly increase due to the proximity to the eddington limit . surprisingly , the mass - loss rate is even quite modest when compared to eta carinae , which has @xmath42 two orders of magnitudes higher than ag car , but a luminosity that is only 3 times higher . \8 . due to the proximity of ag car to the eddington limit and the mass loss , an extended pseudo - photosphere is formed , hiding the hydrostatic surface of the star . even if the mass - loss rate increases towards maximum , that is not sufficient to explain the changes in the spectrum . an expansion in the hydrostatic radius is required in order to reduce the effective temperature . contrary to the current paradigm , the maximum value of the eddington parameter is found during visual minimum , not when the star is evolving towards maximum . therefore , the picture of a quiescent star at visual minimum followed by an eruption during visual maximum due to an increase in the eddington parameter might not correctly describe the s - dor cycles . we thank the anonymous referee for a careful reading of the paper and useful comments . we are grateful to nolan walborn for making the 1989 march dataset available . we acknowledge discussions with jorick vink , stan owocki , georges meynet , cyril georgy , nathan smith , and andre maeder . we are thankful to tom madura and jorick vink for comments on the original manuscript . jhg thanks the max - planck - gesellschaft ( mpg ) for financial support for this work . ad also acknowledge partial financial support from the brazilian agencies fapesp and cnpq . djh gratefully acknowledges partial support for this work from nasa - ltsa grant nag5 - 8211 . this research made use of the smithsonian nasa / ads and simbad ( cds / strasbourg ) databases . , h. j. g. l. m. & de loore , c. w. h. , eds . 1987 , astrophysics and space science library , vol . 136 , instabilities in luminous early type stars ; proceedings of the workshop , lunteren , netherlands , apr . 21 - 24 , 1986	we show that the significantly different effective temperatures ( @xmath0 ) achieved by the luminous blue variable ag carinae during the consecutive visual minima of 19851990 ( @xmath1 k ) and 20002001 ( @xmath2 k ) place the star on different sides of the bistability limit , which occurs in line - driven stellar winds around @xmath3 k. decisive evidence is provided by huge changes in the optical depth of the lyman continuum in the inner wind as @xmath0 changes during the s dor cycle . these changes cause different fe ionization structures in the inner wind . the bistability mechanism is also related to the different wind parameters during visual minima : the wind terminal velocity was 23 times higher and the mass - loss rate roughly two times smaller in 19851990 than in 20002003 . we obtain a projected rotational velocity of @xmath4 during 19851990 which , combined with the high luminosity ( @xmath5 ) , puts ag car extremely close to the eddington limit modified by rotation ( @xmath6 limit ) : for an inclination angle of @xmath7 , @xmath8 for @xmath9 . based on evolutionary models and mass budget , we obtain an initial mass of @xmath10 and a current mass of @xmath11 for ag car . therefore , ag car is close to , if not at , the @xmath6 limit during visual minimum . assuming @xmath12 , we find that @xmath13 decreases from 0.93 to 0.72 as ag car expands toward visual maximum , suggesting that the star is not above the eddington limit during maximum phases .
game theory@xcite is a subject used to predict the strategy of all players in a game . the simplest game is static and non - cooperative game , which describe by payoff function @xmath0 , a linear mapping from strategy space @xmath1 to n - dimension real space @xmath2 . for a game in mixture strategy space , in which every player uses mixture strategy @xmath3 , a probability distribution on strategy set @xmath4 , nash theorem proves that there is always some mixture - strategy equilibrium points , on which no player has the willing to make an independent change . therefor , such equilibrium points can be regarded as a converged points ( or at least fixed points ) of the system , then as the end state of all players . on the other hand , quantum game theory@xcite has been proposed as a quantum version of game theory@xcite . a typical two - player quantum game is defined as @xmath5 , in which @xmath6 is the hilbert space of state of one quantum object , like a photon or electron . such a quantum object plays an important role in quantum game theory . the quantum strategy set @xmath7 or @xmath8 usually is defined as a set of unitary operator on state space @xmath6 , or sometimes subspace of @xmath6 . it s believed that because quantum strategy space is usually larger than the corresponding classical strategy space , one can make use of such advantage of quantum strategy to make money over classical player . generally speaking , a quantum game is not only a quantum version of a classical game . it describes the situation that every player plays game in a strategy space of quantum operators , which transform the state of a quantum object . probably , someday , some original quantum games can be found alone this direction . however , both of classical game theory and quantum game theory are expressed in single player strategy space , so the payoff function is a @xmath9 tensor @xmath10 , mapping a combination of @xmath11 single - player strategies onto a real number . on the contrary , in quantum statistical mechanics , a matrix form of hamiltonian @xmath12 is used in a any - particle case , and the form of density matrix of equilibrium state is always @xmath13 . so a system - level description will unify our formulas for @xmath11-player game , and then maybe improve our understanding and calculation . starting from such an idea , in this paper , we construct a systematical way to reexpress everything into system - level description , including system state and its space , payoff matrix on system space , and reduced single - player payoff matrix . then canonical quantum ensemble distribution is used to describe system equilibrium state . so ideas and technics from statistical physics can be easily applied onto game theory . such application implies a probability that a kinetics equation can be used to describe an evolution that a system ends at the equilibrium state starting from an arbitrary distribution . because the traditional game theory only cares about the macro - equilibrium state , the kinetics equation approach is just pseudo - dynamical equation leading to the equilibrium state . the dynamical process itself might be meaningless . furthermore , besides providing a new pseudo - dynamical approach , the distribution function description has its own meaning . in game theory , maybe general for all economical subjects , usually it s supposed that even the difference between two choices is very small , the high - value one is chosen . this is unnatural when the difference is smaller than the resolution of human decision . therefor , we use a distribution function to replace the maximum - point solution . this means player @xmath14 will choose strategy @xmath15 with probability @xmath16 , even there are another strategy can make more money . here , @xmath17 is the meaning of average resolution level , or in statistical physics , the average noise level . unfortunately , although the ensemble description is the second topic of this paper , only some special case study has been investigated . a general form for any game is still waiting for more exploration . section @xmath18[framework ] constructs the new representation for classical and quantum game . section @xmath18[statphys ] use ensemble distribution and pseudo - dynamical approach to study the equilibrium state in this new representation . discussion of relation between our new representation and quantum , and classical game is included in section @xmath18[framework ] . a lot of questions are pointed out in the discussion section ( @xmath18[discussion ] ) . section @xmath18[conclusion ] is a short summary of the conclusions we have reached . strategy set can be continuous and discrete , and this will effect the mathematical form of all variables , such as the state of player @xmath14 is @xmath19 or @xmath20 , and @xmath0 will be integrations or matrixes . in order to compare with the mathematical form of quantum mechanics and point out the similarity , and to unify classical game theory and quantum game theory , here we use the discrete strategy , although the corresponding transformation of all ideas and formulas is quite straightforward . most of our formulas and results can be generalized into @xmath11-player and @xmath21-strategy game , so for simplicity of expressions , at most time , a @xmath22-player and @xmath23-strategy game is used as our object . for a @xmath11-player game , we suppose the strategy space of player @xmath14 is @xmath24 . the state of player @xmath14 is @xmath25 and @xmath26 . the payoff function of player @xmath14 is a @xmath27-tensor a @xmath11-linear operator , @xmath28 specially , for a 2-player game , @xmath29 can be written as a matrix ( @xmath30-tensor ) so that @xmath31 in which @xmath29 is @xmath32 matrix , not necessary a square one . so a classical game is @xmath33 a general vector in @xmath4 can be defined as @xmath34 in which @xmath35 is the base vector set of strategy space @xmath4 . or in traditional language of game theory , it s a set of all the pure strategies of player @xmath14 . inspired by the application of hilbert space in quantum mechanics , now we introduce two ideas into game theory . first , to redefine the strategy space of single player as a hilbert space . second , to use a system state to replace the single player state . then , at the same time , a new form of payoff function is required to be equivalently defined on the system state . a single - player state vector @xmath36 is written in a new form as @xmath37 it s density matrix form of a mixture state , because a classical strategy of player @xmath14 is to use strategy @xmath15 with probability @xmath38 . the difference between equ([oldsinglestatevector ] ) and equ([newsinglestatevector ] ) can be regarded as just to replace @xmath39 with @xmath40 . the reason of such replacement will be clear when we do it on quantum game . actually , using density matrix to describe mixture state is a approach in quantum mechanics . a system state of all players is defined as @xmath41 a typical form of system state of a @xmath22-player ( player @xmath42 and player @xmath22 ) and @xmath22-strategy ( strategy @xmath43 and strategy @xmath44 ) classical game looks like @xmath45 in fact , the base vector set of hilbert space of system state of @xmath11 players can be defined as direct product of single player base vector as @xmath46 then from equ([newsinglestatevector ] ) and equ([newsystemstatevector ] ) , it can be proved that a system state have the form as @xmath47 in which @xmath48 one can compare this general form with the specific one of @xmath49 game , equ([specificcase ] ) . sometimes , we neglect the subindex and denote @xmath50 as @xmath51 . in such situation , we should notice that a capital @xmath52 denote a general system base vector . such replacement provides a probability to use pure strategy other than the traditional classical pure and mixture strategy . we will discuss this in section @xmath18[quantizedclassical ] . now we try to transform payoff function @xmath29 into system - level form while the invariant condition is equ([singlepayoff ] ) . in a density matrix form , the formula used to calculate the payoff is @xmath53 the solution of equ([singlepayoff ] ) and equ([systempayoff ] ) gives the relation between @xmath54 and @xmath29 . since those two equations should give the same value for any state , we can choose the system state as a pure strategy , or in our language , the base vector of system state . let s denote @xmath55 , which means every player choose a pure strategy @xmath56 , then @xmath57 . then equ([systempayoff ] ) give us @xmath58 so @xmath59 the diagonal elements of @xmath54 can be calculated explicitly . and for our general system density matrix as equ([systemdensity ] ) , only the diagonal terms effect the payoff value @xmath60 , all others can defined as zero . for example , @xmath54 of a 2-player game is @xmath61 this means @xmath54 is diagonal matrix . before we continue our further discussion , let s use one example to present our abstract mathematics and to compare the traditional and new from of state vector and payoff function . the traditional payoff function of prisoner s dilemma is @xmath62 then @xmath63 , g^{2}=\left[\begin{array}{cc } -2 & 0 \\-5 & -4 \end{array}\right].\ ] ] the traditional state vectors are @xmath64 by substituting the above two equations into equ([singlepayoff2 ] ) , we get @xmath65\left[\begin{array}{cc } -2 & -5 \\0 & -4 \end{array}\right]\left[\begin{array}{c}p^{2}_{c}\\p^{2}_{d } \end{array}\right ] \\ \\ & = & -2p^{1}_{c}p^{2}_{c}-5p^{1}_{c}p^{2}_{d}+0\cdot p^{1}_{d}p^{2}_{c}-4p^{1}_{d}p^{2}_{d } \end{array } \label{oldearn}\ ] ] by the new notations , state of player @xmath42 is @xmath66 and system state is @xmath67 or in matrix form , @xmath68.\ ] ] then from equ([relation2 ] ) , we know the new payoff function , @xmath69 , h^{2}=\left[\begin{array}{cccc } -2 & 0 & 0 & 0 \\0 & 0 & 0 & 0 \\0 & 0 & -5 & 0 \\0 & 0 & 0 & -4 \end{array}\right].\ ] ] we can check it by substituting into equ([systempayoff ] ) as , @xmath70 which is the same value with equ([oldearn ] ) . so the new representation includes all the information in the traditional notation , however , more complex it seems . but such complexity brings some other benefit including equilibrium state calculation and generalization into quantum game theory . till now , since the classical strategy is a mixture strategy of the base vector ( strategy ) , we always use density matrix to represent a single player state or a system state , such as in equ([newsinglestatevector ] ) and equ([newsystemstatevector ] ) . now we ask the question that what s the pure state of strategy ( but other than the classical pure strategy ) means in game theory ? a pure strategy vector of player @xmath14 in our representation is @xmath71 therefor , the density matrix of such a pure state is @xmath72 the density matrix of a pure state has off - diagonal elements while the classical mixture density matrix has only the diagonal elements . it seems that pure strategies expand the strategy space . whether it has significant result in game theory or not ? comparing equ([puredensity ] ) with equ([newsinglestatevector ] ) , if we suppose @xmath73 that every diagonal element of pure density equals to corresponding one of mixture density matrix , then those two states will have similar meaning . let s use the prisoner s dilemma as an example again to check if they will give different payoffs . although we still can follow the calculation of mixture state by density matrix method as in equ([systempayoff ] ) , there is an equivalent but much simpler formula for pure state calculation , @xmath74 where @xmath51 is a pure state vector defined direct product of single player state as @xmath75 specially , for prisoner s dilemma , the system state vector is @xmath76 combined with new payoff matrix @xmath77 , @xmath78 \left[\begin{array}{cccc } -2 & 0 & 0 & 0 \\0 & -5 & 0 & 0 \\0 & 0 & 0 & 0 \\0 & 0 & 0 & -4 \end{array}\right]\left [ \begin{array}{c}x^{1}_{c}x^{2}_{c } \\ x^{1}_{c}x^{2}_{d } \\ x^{1}_{d}x^{2}_{c } \\ x^{1}_{d}x^{2}_{d } \end{array}\right ] \\ & = & -2p^{1}_{c}p^{2}_{c}-5p^{1}_{c}p^{2}_{d}+0\cdot p^{1}_{d}p^{2}_{c}-4p^{1}_{d}p^{2}_{d } \end{array}\ ] ] the same value as equ([oldearn ] ) and equ([newearn ] ) . pure state @xmath79 equals to the diagonal term plus some off - diagonal elements . the same payoff value implies that in our situation , only the diagonal term makes sense . in fact , generally , it s because of the diagonal property of @xmath54 . from equ([systempayoff ] ) and equ([relation ] ) , we know @xmath80 it means that even @xmath79 has off - diagonal elements , only the diagonal parts effect the payoff . in one word , the mathematical form of vectors in hilbert space , or the equivalent density matrix form , brings nothing new into game theory but an equivalent mathematical representation . system state can be a pure state or a mixture state , but since the payoff matrix is diagonal , they make no difference . the quantization of classical game is a new game only when _ both density matrix and payoff matrix have off - diagonal elements_. although density matrix @xmath81 can have the off - diagonal terms , the relation of equ([relation ] ) between the new payoff matrix and the traditional one guarantees that @xmath54 can only have the diagonal term . so this quantization condition has no classical meaning in game theory . now the question is if we quantize it anyway , what s the meaning ? is it possible to find any real world objects for such a theory ? if we find such an object , does the relative phase in state vector play any roles in such situation ? and further more , the vector space gives us the freedom to choose our base vectors , does such transformation play any roles ? the expanded strategy space provides another class of possible state . in the mixture classical density matrix , a system density matrix ia always has the form as equ([systemdensity ] ) , which is a direct product of all single players . if pure state is permitted , a general system density matrix may not be a direct product , but an entangled state of all single players . does such an entangled density matrix have significant effect on game theory ? at the last of this section , lets come back to question of the meaning of the off - diagonal term of payoff matrix @xmath54 by referring to an example . what s the meaning of @xmath82 in the payoff matrix below ? @xmath83 , h^{2}=\left[\begin{array}{cccc } -2 & 0 & 0 & \epsilon_{2 } \\0 & 0 & 0 & 0 \\0 & 0 & -5 & 0 \\\epsilon_{1 } & 0 & 0 & -4 \end{array}\right].\ ] ] the representation of classical game above strongly depends on the base vectors of strategy set . but in classical game , such base vectors are predetermined and artificial . they are just the original discrete basic classical strategies . no inner product has been predefined between them before our construction of the new representation . now we turn to quantum game theory , and fortunately it will provide us a very natural explanation of our base vectors . the proposed and developing quantum game theory is different with our quantized classical game theory . while our approach is a representation , the quantum game theory is a quantum version of game theory . it use the idea of game theory , but all operations ( strategies ) and the object of such operations are from quantum world@xcite . a typical 2-player quantum game is defined by @xmath84 in which @xmath6 is the hilbert space of the state of a quantum object , @xmath85 is the initial state of such an object , @xmath86 is player @xmath87 or @xmath88 s set of quantum operators acting on @xmath6 . and @xmath89 are their payoff functions . using well - known quantum penny flip game@xcite as an example , spin of an electron is used as penny , so the base vectors of @xmath12 are @xmath90 . the initial state is chosen as @xmath91 . the classical operators are to flip the penny or not , so they are @xmath92 , f = \left[\begin{array}{cc}0 & 1\\1 & 0\end{array}\right].\ ] ] the quantum operator can be a general unitary operator @xmath93.\ ] ] the payoff function is usually defined as @xmath94 in which @xmath95 = -p_{b}\ ] ] means , player @xmath87 gets @xmath42 when the quantum object ends in @xmath96 state and loses @xmath42 when in @xmath97 state . considering the relation between classical game and quantum game , we redefined quantum game with a slightly difference with the definition equ([qgame ] ) as @xmath98 in which @xmath99 is the set of quantum operators while @xmath100 is the set of classical operators , usually @xmath100 is a subset of @xmath99 , but not necessary . now let s use our new mathematical representation to reexpress the quantum game theory . because quantum game theory is constructed on the basis of quantum state of a quantum object , @xmath6 , it provides a set of natural base vectors of strategy space . for quantum penny flip game , all strategy are operators with the form of @xmath101 so the base vectors are @xmath102 . furthermore , if we define the inner product of operator as @xmath103 @xmath104 is a complete orthogonal base vector set for all quantum operators . then the operators can be regarded as vectors in hilbert space , a hilbert space of operator , which we denote as @xmath105 . the classical strategies are @xmath106 in order to form another complete orthogonal base set , we need to define other two base vectors as @xmath107 then operator space can also be expressed by @xmath108 , while the classical strategy is @xmath109 . @xmath110 is the operator to turn the @xmath96 into @xmath96 , no definition when the initial state is @xmath97 . a meaningful operator should give the end result for starting state both as @xmath96 and @xmath97 , so operators @xmath111 are better than @xmath112 in this . another advantage is that all the base vector are unitary and hermitian operator . it s easy to prove that under our definition of inner product , the matrix form of @xmath113 is just @xmath114 . for such base strategies , @xmath115 , so @xmath116 and @xmath113 have the same the matrix forms . but @xmath117 . so a unitary operator is not in unit magnitude . one way to solve such conflict is redefine base vectors as @xmath118 , but here we prefer another way to keep the form of unitary operator , and to redefined inner product as @xmath119 applying our representation onto this quantum penny flip game , the system state is @xmath120 the single - player state coming from classical sub strategy space is @xmath121 if we quantize it anyway as we did in section @xmath18[quantizedclassical ] that in a pure state of quantized classical game , the single - player state is @xmath122 then @xmath123 will have off - diagonal term , while @xmath124 only has the diagonal term . now , applying our representation onto quantum strategy space of this penny flip game , the single - player state is @xmath125 here @xmath126 can be any one of @xmath127 . next step , we allow our strategy can be mixture state in @xmath104 . the single - player state of a general mixture strategy will be @xmath128 where @xmath129 are quantum pure strategies , may or may not equal to @xmath130 . this means @xmath131 is mixture state but maybe diagonal in other set of base vectors . a more general system state can be constructed in the quantum strategy space @xmath104 by destroying equ([quantumdensity ] ) . we ever mentioned in the last part of section @xmath18[quantizedclassical ] that density matrix of a general state is not required to be a direct product to density matrix of every single player . but still , the meaning of such state is not clear here . in the later discussion , we name equ([c ] ) , equ([q ] ) , equ([q ] ) and equ([qq ] ) as classical game ( cg ) , quantized classical game ( qcg ) , pure - strategy quantum game ( pqg ) and quantum game ( qg ) , respectively . and in the classical game , when the system density matrix is not a direct product , we call the game as entangled classical game ( ecg ) , while for quantum case , entangled quantum game ( eqg ) . the strategy space of all these games have the relation that @xmath132 the relation is ` @xmath133 ' not ` @xmath134 ' because it s possible that the later has no independent meaning other than the former although the later has a larger strategy space . in fact , for classical game , we still have another smaller strategy space classical pure strategy space , and the game in that space pure - strategy classical game ( pcg ) . fortunately , the relation between pcg and cg is already clear enough through nash theorem , so it s not necessary to discuss it anymore . now all games have been unified in our mathematical representation . everyone has its own strategy space and base strategy vectors . in order to finish presenting our representation , we need to calculate the new payoff function @xmath54 . let s still use the quantum penny flip game as an example . in quantum game , because the non - commutative relation between operators ( base vectors ) , the order of acting effects the results . on the contrary , in classical game , usually the base vectors are commutative , so the order of acting does nt matter . we can see this by @xmath135 = 0,\ ] ] but @xmath136 = \left[\begin{array}{cc}0 & -2i\\-2i & 0 \end{array}\right ] = -2if^{c } \neq 0.\ ] ] we define the order is @xmath137 , then the original payoff function @xmath29 is defined to take the value of @xmath138 where @xmath139 is anyone of @xmath140 . this will give all @xmath141 values of @xmath142 . in classical game , we require and notice that @xmath29 is naturally a @xmath30-tensor , because the payoff of mixture strategy is the weighted average with their own probability . the linear property of this payoff is a requirement of our new system - level payoff function , which is a @xmath143-tensor , or we say , linear for right vector , anti - linear for left vector . but , here in quantum game , from equ([quantumsinglepayoff ] ) we find @xmath29 is definitely not a tensor , not a linear mapping , @xmath144 then is it possible to transform such payoff into system - level @xmath143-tensor ? we need to prove it . from classical game , one thing we already know that in the classical state subspace @xmath145 , no matter the strategy is pure or mixture , such transformation exists . so we use three steps to prove that a system - level @xmath143 tensor payoff can be constructed . first , for system state only staying on one base vectors @xmath146 , so that @xmath147 we define the elements @xmath148 for example , @xmath149 , which means when both player @xmath42 and player @xmath22 choose @xmath150 , player @xmath42 wins ; @xmath151 , which has no classical meaning , because it s a off - diagonal elements . it s easy to prove that @xmath54 is hermitian , @xmath152^ { * } \\ & = & \left[tr\left(s^{2'}s^{1'}\rho^{\dag}_{0}s^{1}s^{2}\left(p^{i}\right)^{\dag}\right)\right]^ { * } \\ & = & \left[tr\left(p^{i}s^{2'}s^{1'}\rho^{\dag}_{0}s^{1}s^{2}\right)\right]^ { * } \\ & = & \left(h^{i}_{ss^{'}}\right)^ { * } , \end{array}\ ] ] in which , we require @xmath153 . so @xmath154 second , we prove that for a system state not staying on the base vector , but on pure state , such as @xmath155 we still have @xmath156 , in which @xmath54 is a @xmath157-tensor . * proof * : from payoff definition equ([quantumsinglepayoff ] ) , @xmath158 in which we need the property that @xmath159 is a linear operator . on the other hand , when @xmath54 is a @xmath143-tensor , @xmath160 using the definition of @xmath161 in equ([quantumpayoffmatrix ] ) , we know they equal . at last , we prove for a mixture state , such as @xmath162 we still have @xmath163 . * proof * : using above result , @xmath164 therefor , for any system state we still have @xmath165 the payoff matrix of the quantum penny flip game is a @xmath166 matrix @xmath167\ ] ] and @xmath168 . compare with the payoff matrix of classical game with the payoff matrix of quantum game , a significant difference is that the later has non - zero off - diagonal elements while the former only has diagonal elements . through this representation we know the difference between classical game and quantum game is not only the size of strategy space but also the off - diagonal elements of payoff matrix . from @xmath54 above , the sub - matrix related with @xmath169 is @xmath170= -h^{2,c}.\ ] ] they are different with the new payoff matrix of the original classical game , @xmath171= -h^{2,c}_{original},\ ] ] which only has diagonal terms . so we can say , the quantization process changes the definition of the original classical game . however , the meaning of quantum game does not completely depends on the equivalence with the original classical game . so even given such difference , quantum game is still probably a new game . if we defined a quantum game by payoff matrix @xmath54 , another privilege of this new representation is that the definition of a quantum game is independent on @xmath172 , the state of a quantum object . our payoff function can be directly defined on system state @xmath173 . of course , any payoff defined on @xmath172 can be transferred equivalently into a function on @xmath173 . so a quantum game can be defined as @xmath174 in which @xmath175 has base vectors @xmath176 , and @xmath100 has base vectors @xmath177 . usually the later is a subset of the former . a classical payoff function is defined on system base vectors such as @xmath178 , while a quantum payoff function is defined as @xmath179 . if @xmath100 is a subspace of @xmath99 , the payoff matrix of classical game just take the corresponding diagonal parts whether the corresponding sub - matrix of the quantum game payoff matrix has non - zero off - diagonal terms or not . in the next part of this paper , we try to make use of this representation and hopefully to find something significant . in the new representation , a game seems very similar with an ising model with global interaction . the payoff of every player is related with everyone else . the state of every player can be represented by a quantum state vector or density matrix . every player try to stay at point with the maximum payoff , while in ising model , the whole system try to stay at minimum energy point . the distribution of a quantum system at thermal equilibrium is @xmath180 where @xmath12 is the system hamiltonian , @xmath181 is so called partition function . now as in statistical mechanics , we introduce the idea of distribution function of state into game theory , but instead of function in @xmath182 space in statistical mechanics , here in @xmath183 space , the state space of every single player . a natural form is @xmath184 in which @xmath185 is the payoff function of player @xmath14 in its own strategy space and @xmath181 is the partition function in @xmath14 s strategy space . the payoff matrix @xmath54 we have now is defined in system strategy space . so a kind of reduced matrix is what we need to find . before the detailed calculation , one thing we should notice that the equilibrium density matrix description is different with the classical mixture strategy . if the eigenvectors of @xmath185 can be found as @xmath186 , then @xmath187 is similar with the classical mixture strategy form , and @xmath188 can be regarded as the probability on strategy @xmath189 . but first , such an set of eigenvectors is not always the same as the classical base vectors , because sometimes , we have non - zero off - diagonal elements . second , such a density matrix gives the probability of any pure strategies @xmath190 even being different with the base vector , by @xmath191 this is impossible in mixture strategy description . now we start to define the reduced payoff matrix and investigate its properties . a nash equilibrium state is defined that at that point every player is at the maximum point due to the choices of all other players are fixed . a reduced payoff matrix should describe the payoff of a single person when the choice of all other players . in the traditional language of game theory , such a reduced payoff matrix is equivalently to be defined like the end result of equ([oldearn ] ) under any arbitrary fixed @xmath192 . but we need a matrix form here . for pure system strategy , @xmath193 is a @xmath143-tensor . in a 2-player game , @xmath194 can also be regarded as a @xmath195-tensor . a reduced payoff matrix of player @xmath14 means in the viewpoint of player @xmath14 it should be a @xmath143-tensor . when both player @xmath42 and player @xmath22 stays on pure strategy @xmath196 respectively , it has a natural definition , @xmath197 . since our strategy can be a mixture state , or generally a density matrix form , we need to generalize the above definition . a reduced payoff matrix of player @xmath42 in a @xmath22-player game is defined as @xmath198 where @xmath199 is the trace in subsapce of player @xmath22 . from equ([payoffvalue ] ) , the payoff value of player @xmath42 is @xmath200 so if we know the reduced payoff matrix of player @xmath42 , the payoff value can be calculated by @xmath201 in fact the @xmath199 action is quite hard to perform , because this requires the result of a trace is a matrix , not a number as usual . an equivalent but easily understood form of equ([reducedpayoff ] ) is @xmath202 in which @xmath203 is a sub matrix with fixed player @xmath42 s index ( here , first and third index ) . in order to define a general form for @xmath11-player game , we denote the trace @xmath204 as diagonal summation in the space except player @xmath14 s . so in @xmath22-player game , @xmath205 . then a general reduced payoff matrix of player @xmath14 under fixed strategies of all other players is @xmath206 still using the prisoner s dilemma as example , when player @xmath22 choose strategy @xmath207 with @xmath208 and @xmath209 with @xmath210 , the state is @xmath211\ ] ] then @xmath212\left[\begin{array}{cc\|cc } -2 & 0 & 0 & 0 \\0 & -5 & 0 & 0 \\ \hline 0 & 0 & 0 & 0 \\0 & 0 & 0 & -4 \end{array}\right]\right ) \\ & = & \left [ \begin{array}{cc } -2p^{2}_{c}-5p^{2}_{d } & 0 \\0 & 0\cdot p^{2}_{c}-4p^{2}_d \end{array}\right ] . \end{array}\ ] ] recalls metropolis method and its derivative heat bath@xcite method in monte carlo simulation of statistical ensemble . in the simulation of equilibrium state of ising model , every single step , when a random spin is chosen , it faces the same situation with our game player . all other spins have decided one state to stay temporary , it has some choices of its own state by evaluating the energy difference between all its possible states . then it choose one state to stay by a transition probability or transition rate over all possible states . the kinetics equation for such process is not unique , different forms of transition probability can give the same equilibrium state . now we face a quantum system , although a similar situation . every player should make his decision every step with the fixed state of all other players and we also ask for the equilibrium state . the reason that different kinetics equations give the same equilibrium state in statistical physics is the well - known detailed balanced theorem in thermal equilibrium , but we do nt have a corresponding one in game theory . we now just suppose that at equilibrium state , the density matrix of player @xmath14 s state is @xmath213 and we choose a heuristic kinetics equation as iteration equation , @xmath214 then the equilibrium state is defined as the fixed point of this iteration if it has fixed point . in fact , kinetics equation equ([kineticsequation ] ) is @xmath11 related iteration equations . the existence of the fixed point is not obvious . even the questions itself is not unique , although the experience in simulation in statistical physics implies that such equation should exist probably with different form . the fixed point might be different with nash equilibrium even if it exists . in this paper , all these questions are neglected . let s first test such idea in some examples , just like what a physicist usually does , not a mathematician , who will pay more attention on a general definition of equilibrium state and the proof of the existence . equ([kineticsequation ] ) of a classical game is much easier to deal with than the one of a quantum game . in classical game , both @xmath54 and @xmath185 are diagonal . the density matrix at time @xmath215 can always be written as @xmath216 , then equ([kineticsequation ] ) will lead to a series of evolution equations for @xmath217 . however , in quantum game , since the payoff matrix @xmath54 has off - diagonal elements , the reduced payoff matrix @xmath185 also can have off - diagonal elements . then the density matrix can be equivalently replaced by evolution equation of @xmath218 * only * when the density matrix is expressed in the base vector formed by the eigenvectors of @xmath185 . but with off - diagonal elements , such eigenvectors are not always the base vector we used to express the game and they might change during the iteration process . so the first step is to solve the eigenvalue equation of @xmath54 and @xmath185 . the eigenvalue problem in classical game is quite easy . all the eigenvectors are the base vector we used , the eigenvalues are just the corresponding diagonal elements . in a quantum game , it depends on the details of payoff matrix . for example , in the quantum penny flip game , the payoff matrix @xmath77 has @xmath166 elements . even when player @xmath22 choose @xmath219 , the reduced payoff matrix of player @xmath42 is @xmath220\ ] ] the eigenvalues and the corresponding eigenvectors are @xmath221^{t } \\ 2 & \rightarrow & \left[1,0,1,0\right]^{t } \\ 0 & \rightarrow & \left[-1,0,1,0\right]^{t } \\ 0 & \rightarrow & \left[0,i,0,1\right]^{t } \end{array } \right).\ ] ] this means a quantum player can make money over the classical player with @xmath222 by using strategy @xmath223^{t}$ ] . and the funny thing is the value of payoff the 2 , not @xmath42 in classical case when player @xmath42 uses @xmath224^{t}$ ] . it clearly shows the effect of the off - diagonal elements for a quantum player . if the player @xmath42 is still a classical player , the strategy he can use is just @xmath225 , so he will get @xmath226 respectively . anyway , the topic of this section is show the way to do the iteration defined in the kinetics equation equ([kineticsequation ] ) , not the reason of such difference . now we have the idea . starting , for instance , from player @xmath22 choose @xmath227 , the state of player @xmath42 is then @xmath228 and then substitute it back to equ([kineticsequation ] ) and do the iteration . however , from above density matrix we know that even beginning from a pure state , the state after one iteration will be a mixture state . in classical game , it does nt matter , because the end state generally can be a mixture state , and a pure state is equivalent with a mixture state with the same diagonal part . but in quantum game , mixture strategy is quite different with a pure one . one way to deal with this problem is to set @xmath229 . then the kinetics equation of quantum game becomes , @xmath230 in which @xmath231 is the eigenvector with maximum eigenvalue of the reduced payoff matrix . so only the maximum one is kept after every step . but this will brings new problems when the real equilibrium state is a mixture state . so for quantum game , it s better to regard the approach shown here just as an idea . later on , classical game is our main object . the task of this section is just to point out that a quantum game brings new things such as eigenvalue problem while which is quite trivial in classical game . now we come to use our kinetics equation approach on some examples of classical game . first , let s finish the discussion about the prisoner s dilemma . we already know @xmath232 \end{array } & \mbox{and } & \begin{array}{lll } h_{r}^{2 } & = & \left [ \begin{array}{cc } -2p^{1}_{c}-5p^{1}_{d } & 0 \\0 & 0\cdot p^{1}_{c}-4p^{1}_d \end{array}\right ] . \end{array } \end{array}\ ] ] suppose we start from player @xmath22 with @xmath233 , then from equ([kineticsequation ] ) , @xmath234 when @xmath229 , which means infinite resolution level , or we say , any difference in payoff is significant , then @xmath235 . the equilibrium state is @xmath236 , which is equivalent with nash equilibrium . when @xmath17 is finite , denote the fixed point as @xmath237 . the stability of this fixed point can be analyzed by the linear stability matrix , @xmath238_{\left(p^{1,}_{c},p^{2,}_{c}\right ) } \triangleq \left(\frac{\partial p^{i}_{\alpha}}{\partial p^{j}_{\mu}}\right)_{\left(\prod_{i}l_{i}\right)\times\left(\prod_{i}l_{i}\right)}\left\|_{\left(p^{i,}_{\nu}\right)}\right.\ ] ] in this specific case , it s unstable , the fixed point graph of the kinetics equation is shown in fig([graphprisoner ] ) . when @xmath239 , which means the players care nothing about the payoff , then @xmath240 . of course , such solution is useless , but still consistent with our intuitive result . growths . because here only one parameter @xmath241 we need to calculate , a simple fixed point graph shows the result . usually in a multi - strategy game or with more players , we will have more parameters and more complex equations . then in that situation , we will have to use simulation . the function plotted here is @xmath242 . because the two steps of one iteration use the same function form , it can be regarded as two iteration steps with only one function . ] second example , we choose hawk - dove , a two - pure - ne game . the payoff matrix of player @xmath42 and @xmath22 are @xmath243 , h^{2 } = \left [ \begin{array}{cccc}3 & 0 & 0 & 0 \\0 & 4 & 0 & 0 \\0 & 0 & 1 & 0 \\0 & 0 & 0 & 0 \end{array } \right]\ ] ] the reduced payoff matrix is @xmath244\ ] ] then the kinetics equation is @xmath245 it s easy to know that when @xmath229 , fixed point are @xmath246 , @xmath247 and @xmath248 depending the initial state @xmath249 , @xmath250 or @xmath251 . for a finite @xmath17 , a fixed point graph is shown in fig([graphhawk ] ) . . there are three fixed points for this iteration process . one is unstable fixed point , @xmath252 , while the other two are stable fixed points closing to @xmath253 and @xmath42 respectively , when @xmath17 grows . in figure ( b ) , we explicitly show the two iterations instead of the self - mapping from @xmath254 to @xmath254 in figure ( a ) . it gives more details . @xmath255 is still an unstable fixed point . while the end state can be a @xmath256 or @xmath257 depending on the initial state . ] the third example is the classical penny flip game , which has no pure ne . the payoff matrix are @xmath258 = -h^{2}.\ ] ] the reduced payoff matrix is @xmath259 , h^{2}_{r } = \left [ \begin{array}{cc}p^{1}_{f}-p^{1}_{n } & 0 \\ 0 & p^{1}_{n}-p^{1}_{f } \end{array } \right].\ ] ] then the kinetics equation is @xmath260 from fig([graphpenny ] ) , @xmath261 is the only fixed point no matter @xmath262 or not . and even the only fixed point is unstable . . the only fixed but unstable point is always @xmath252 for different @xmath17 . in figure ( b ) , two iteration functions are used to show a clearer but more complex picture , from which we know @xmath263 is still a fixed point , but if starts from initial state other than this point , the system will jump between @xmath264 , @xmath256 , @xmath143 and @xmath257 . ] the kinetics equation , its fixed points and the stability analysis of the fixed points gives a method to find equilibrium state and to refine them if we require an applicable equilibrium point is a stable fixed point . from all the trivial cases we tested it seems valid . but questions such as more tests , a general form of such equation , and the relation between such fixed points and nash equilibrium is waiting for more detailed discussion . at last , we have to admit that our simulation is not equivalent with the kinetics equation . pure strategies are included by the kinetics equation , but since our algorithm is classical , here we only let it evolute in the subspace of mixture strategy . for classical game , this is not a fatal problem , because we have prove that pure strategy is equivalent with mixture strategy having the same diagonal part . but for a quantum game , pure strategy is totaly different with the mixture classical one because of the off - diagonal elements of payoff matrix . is it possible to find such a simulation algorithm ? on the other hand , when @xmath265 , the fixed point of our kinetics equation might not equal to the nash equilibrium state . such fixed points are the end states when the average resolution level of all players is @xmath17 , which can be regarded as a typical scale which players care . this concepts may expand the description of game theory into the situation that players are not complete rational . they can evaluate the payoff , but not explicitly , only a rough range . and from the experience in statistical physics , especially phase transition , we know that even when @xmath17 is not very large but large enough the lowest energy mode ( here , maximum payoff mode ) will dominate the system . this means , under some not extremely restricted conditions , the traditional nash equilibrium is still valid . it will be funny if one can prove such conclusion from a general situation in our equilibrium definition . it s quite straightforward to extend our notation into @xmath11-player game and continuous strategy case . however , although a new representation has been introduced to express everything in a static game , the advantage of such a language and the meaning of all other games is still open . and further more , if it s acceptable for static game , is it possible to be developed into evolutionary game theory ? and cooperative game ? is it related with entangled system state ? as discussed in section [ phase ] , because of the iteration procedure and the distribution function we used , a natural way of equilibrium calculation and refinement is provided by our pseudo - dynamical method . the non - trivial phase transition happening in statistical physics at finite @xmath266 implies the probability that when @xmath267 the traditional equilibrium state can be reached at some finite noise level , not necessary at no - noise infinite - resolution background . in this paper , we only argued such possibility , not by a real example . further analysis should be done to confirm such statement , although we believe it from the background in physics . and as discussed in section [ qqgame ] , when our representation is used in quantum game theory , a set of base vectors of strategy ( operator ) space and their inner product need to be defined to form them as a hilbert space . then all the other procedures are quite straightforward . at least , it gives equivalent description . but there are still some open questions , like what s the meaning of a non - unitary operator in the strategy space ? does physical operator have to be unitary operator ? another interesting question is the effect of base vector transformation of hilbert space . what happens if base vectors other than our @xmath268 are used ? we have to say our present result is a theory far from complete . it stacks in our hands for a very long time , now we want to share the idea with all . in fact , it s even possible to be nothing than a toy representation of game theory . however , even in such case , it s still of little value to provide a unified description and a possible pseudo - dynamical equation theory which might be completed later so that the end state of iteration from an arbitrary initial will be the equilibrium state . as you may already noticed our paper is filled with questions other than their answers . hopefully it will motivate the discussion . ironically , during the revision of this paper , we found that the idea using a hilbert space to describe classical and quantum strategies has been proposed in @xcite long time before . so our works can be regarded as a realization and development of this idea . in our paper , not only strategies , but also payoff functions has been reexpressed into hilbert space and operators on it . besides lots of questions in above section , here we summarize the reliable conclusions we have till now . first , in the new representation , all games including classical , quantum , even entangled game , under general @xmath11-player @xmath269 case , can be defined by a unified definition as equ([ourquantumgame ] ) . all the difference among the games is at the base vectors of strategy space and the payoff matrix a @xmath143-tensor . in the traditional form , payoff function of @xmath11-player classical game is @xmath270-tensor ; and for quantum game , it depends on @xmath6 and @xmath85 , even not a tensor form . we have to use special language for every specific game . second , in our representation , for quantum games , it s easy to see the role of off - diagonal elements of payoff matrix when we say a quantum player can make more money over classical player . if the payoff matrix is diagonal , it makes no difference , although a quantum player can make use of quantum pure strategy , which has off - diagonal elements in density matrix . game quantum is only possible when both density matrix and payoff matrix have off - diagonal elements . however , unfortunately , as we have pointed out in section @xmath271 , quantization of classical game probably changes the definition of original classical game . third , with the form of payoff matrix and reduced payoff matrix , equilibrium density matrix in quantum statistical mechanics @xmath272 gives an equilibrium distribution over strategy space . this provides some flexibility on the application of game theory such as average behavior and collapse into nash equilibrium under infinite resolution level ( @xmath229 ) . this helps take partial rationality into our theory . therefor , for classical game , although our new representation brings nothing new theoretically , it may provide some new technical tools to analyze the ne , to include partial rationality , and to develop evolutionary game . if such a representation can provide some other insightful advantage besides an equivalent representation of both classical and quantum game theory , it s necessary to try more real games , both classical and quantum , in the new framework . from the section @xmath18[eigenvalue ] , we see that because the system space is the direct space of all players , the matrix form will be so large that it make all calculations un - convenient . in quantum mechanics , the idea to solve such problem is to introduce particle - number representation to replace direct product of base vectors . for undistinguishable particles , such approach significantly reduce the hardwork of calculation . maybe such simplification can be generalized into game theory . the author want to thank dr . shouyong pei , qiang yuan , zengru di , yougui wang and dahui wang for their inspiring discussion and the great patience to listen to my struggling argument at anytime . thanks also is given to prof . zhanru yang , dr . j. eisert and dr . l. marinatto for their suggestions to the revision . 99 m. j. osborne and a. rubinstein , _ a course in game theory _ , mit press , 1994 . friedman , _ game theory with application in economics _ , new york : oxford university press , 1990 . e.w . piotrowski and j. sladkowski , an invitation to quantum game theory , int . 42(2003 ) , 1089 . flitney and d. abbot , an introduction to quantum game theory , fluctuation and noise letters vol . * 2 * , no . * 4(2002 ) , r175-r187 . j. eisert , m. wilkens , and m. lewenstein , quantum games and quantum strategies , phys . lett , 83(1999 ) * , 3077 . meyer , quantum strategies , phys . rev . lett . * 82(1999 ) , 1052 . l. marinatto and t. weber , a quantum approach to static games of complete information , phys . a 272*(2000 ) , 291 . newman and g.t . barkema , _ monte carlo method in statistical physics _ , new york : oxford university press , 1999 .	in this paper , we introduce a framework of new mathematical representation of game theory , including static classical game and static quantum game . the idea is to find a set of base vectors in every single - player strategy space and to define their inner product so as to form them as a hilbert space , and then form a hilbert space of system state . basic ideas , concepts and formulas in game theory have been reexpressed in such a space of system state . this space provides more possible strategies than traditional classical game and traditional quantum game . so besides those two games , more games have been defined in different strategy spaces . all the games have been unified in the new representation and their relation has been discussed . general nash equilibrium for all the games has been proposed but without a general proof of the existence . besides the theoretical description , ideas and technics from statistical physics , such as kinetics equation and thermal equilibrium can be easily incorporated into game theory through such a representation . this incorporation gives an endogenous method for refinement of equilibrium state and some hits to simplify the calculation of equilibrium state . the more privileges of this new representation depends on further application on more theoretical and real games . here , almost all ideas and conclusions are shown by examples and argument , while , we wish , lately , we can give mathematical proof for most results . key words : game theory , quantum game theory , quantum mechanics , statistical physics pacs : 02.50.le , 03.67.-a , 03.65.yz , 05.20.-y , 05.30.-d
multi - dimensional or multi - way data is prevalent nowadays , which can be represented by tensors . an @xmath0th - order tensor is a multi - way array of size @xmath1 , where the @xmath2th dimension or mode is of size @xmath3 . for example , a tensor can be induced by the discretization of a multivariate function @xcite . given a multivariate function @xmath4 defined on a domain @xmath5^n$ ] , we can get a tensor with entries containing the function values at grid points . for another example , we can obtain tensors based on observed data @xcite . we can collect and integrate measurements from different modalities by neuroimaging technologies such as functional magnetic resonance imaging ( fmri ) and electroencephalography ( eeg ) : subjects , time , frequency , electrodes , task conditions , trials , and so on . furthermore , high - order tensors can be created by a process called tensorization or quantization @xcite , by which a large - scale vectors and matrices are reshaped into higher - order tensors . however , it is impossible to store a high - order tensor because the number of entries , @xmath6 when @xmath7 , grows exponentially as the order @xmath0 increases . this is called the `` curse - of - dimensionality '' . even for @xmath8 , with @xmath9 we obtain @xmath10 entries . such a huge storage and computational costs required for high dimensional problems prohibit the use of standard numerical algorithms . to make high dimensional problems tractable , there were developed approximation methods including sparse grids @xcite and low - rank tensor approximations @xcite . in this paper , we focus on the latter approach , where computational operations are performed on tensor formats , i.e. , low - parametric representations of tensors . in this paper , we consider several tensor formats , especially the tensor train ( tt ) format , which is one of the simplest tensor networks developed with the aim of overcoming the curse - of - dimensionality . extensive overviews of the modern low - rank tensor approximation techniques are presented in @xcite . the tt format is equivalent to the matrix product states ( mps ) for open boundary conditions proposed in computational physics , and it has taken a key role in density matrix renormalization group ( dmrg ) methods for simulating quantum many - body systems @xcite . it was later re - discovered in numerical analysis community @xcite . the tt - based numerical algorithms can accomplish algorithmic stability and adaptive determination of ranks by employing the singular value decomposition ( svd ) @xcite . its scope of application is quickly expanding for addressing high - dimensional problems such as multi - dimensional integrals , stochastic and parametric pdes , computational finance , and machine learning @xcite . on the other hand , a comprehensive survey on traditional low - rank tensor approximation techniques for cp and tucker formats is presented in @xcite . despite the large interest in high - order tensors in tt format , mathematical representations of the tt tensors are usually limited to the representations based on scalar operations on matrices and vectors , which leads to complex and tedious index notation in the tensor calculus . for example , a tt tensor is defined by each entry represented as products of matrices @xcite . on the other hand , representations of traditional low - rank tensor formats have been developed based on multilinear operations such as the kronecker product , khatri - rao product , hadamard product , and mode-@xmath2 multilinear product @xcite , which enables coordinate - free notation . through the utilization of the multilinear operations , the traditional tensor formats expanded the area of application to chemometrics , signal processing , numerical linear algebra , computer vision , data mining , graph analysis , and neuroscience @xcite . in this work , we develop extended definitions of multilinear operations on tensors . based on the tensor operations , we provide a number of new and useful representations of the tt format . we also provide graphical representations of the tt format , motivated by @xcite , which are helpful in understanding the underlying principles and tt - based numerical algorithms . based on the tt representations of large - scale vectors and matrices , we show that the basic numerical operations such as the addition , contraction , matrix - vector product , and quadratic form are conveniently described by the suggested representations . we demonstrate the usefulness of the proposed tensor operations in tensor calculus by giving a proof of orthonormality of the so - called frame matrices . moreover , we derive explicit representations of localized linear maps in tt format that have been implicitly presented in matrix forms in the literature in the context of alternating linear scheme ( als ) for solving various optimization problems . the suggested mathematical operations and tt representations can be further applied to describing tt - based numerical methods such as the solutions to large - scale systems of linear equations and eigenvalue problems @xcite . this paper is organized as follows . in section 2 , we introduce notations for tensors and definitions for tensor operations . in section 3 , we provide the mathematical and graphical representations of the tt format . we also review mathematical properties the tt format as a low - rank approximation . in section 4 , we describe basic numerical operations on tensors in tt format such as the addition , hadamard product , matrix - vector multiplication , and quadratic form in terms of the multilinear operations and tt representations . discussion and conclusions are given in section 5 . the notations in this paper follow the convention provided by @xcite . table [ table : notation_ten ] summarizes the notations for tensors . scalars , vectors , and matrices are denoted by lowercase , lowercase bold , and uppercase bold letters @xmath11 , @xmath12 , and @xmath13 , respectively . tensors are denoted by underlined uppercase bold letters @xmath14 . the @xmath15th entry of @xmath14 of size @xmath16 is denoted by @xmath17 or @xmath18 . a subtensor of @xmath14 obtained by fixing the indices @xmath19 is denoted by @xmath20 or @xmath21 . we may omit ` : ' as @xmath22 if the rest of the indices are clear to readers . the mode-@xmath2 matricization of @xmath23 is denoted by @xmath24 . we denote the mode-@xmath25 matricization of @xmath14 by @xmath26)}\in{\mathbb{r}}^{i_1i_2\cdots i_n\times i_{n+1}\cdots i_n}$ ] in the sense that @xmath27\equiv \{1,2,\ldots , n\}$ ] is the set of integers from 1 to @xmath2 . in addition , we define the multi - index notation by @xmath28 for @xmath29 @xmath30 . by using this notation , we can write an entry of a kronecker product as @xmath31 . moreover , it is important to note that in this paper the vectorization and matricization are defined in accordance with the multi - index notation . that is , for @xmath23 , we have @xmath32 ) } \in{\mathbb{r}}^{i_1\cdots i_n \times i_{n+1}\cdots i_n } \quad & \leftrightarrow \quad { \mathbf{x}}(\overline{i_1\cdots i_n},\overline{i_{n+1}\cdots i_n } ) = { \underline{\mathbf{x}}}(i_1,i_2,\ldots , i_n ) , \end{split}\ ] ] for @xmath30 . .[table : notation_ten]notations for tensors [ cols="<,<",options="header " , ] let @xmath33 and @xmath34 be tt tensors . the sum @xmath35 can be expressed in the tt format @xmath36 that is , each tt - core of the sum is written as the direct sum of the tt - cores . alternatively , each entry of @xmath37 can be represented as products of matrices @xmath38 the tt - ranks for @xmath39 are the sums , @xmath40 . on the other hand , multiplication of @xmath14 with a scalar @xmath41 can be obtained by simply multiplying one core , e.g. , @xmath42 , with @xmath43 as @xmath44 . this does not increase the tt - ranks . we note that that the set of tensors with tt - ranks bounded by @xmath45 is not convex , since a linear combination @xmath46 generally increases the tt - ranks , which may exceed @xmath45 . the hadamard ( elementwise ) product @xmath47 of @xmath48 and @xmath49 can be written in the tt format as @xmath50 that is , each tt - core is written as the mode-@xmath51 kronecker product . as an alternative representation , each entry is written as products of matrices @xmath52 the tt - ranks for the hadamard product are the multiplications , @xmath53 . the contraction of two tensors @xmath54 and @xmath55 is defined by @xmath56 the contraction of a tt tensor @xmath33 with a rank - one tensor @xmath57 can be simiplified as @xmath58 the contraction of two tt tensors @xmath59 and @xmath60 can be calculated by combining the hadamard product and the contraction with the rank one tensor @xmath61 as @xmath62 where @xmath63 for @xmath30 . we remark that @xmath64 and @xmath65 are row and column vectors . we define a generalized contraction operator of two tt - cores as follows . the core contraction of two tt - cores @xmath66 and @xmath67 is defined by @xmath68 we can express the contraction of two tt - tensors @xmath33 and @xmath34 by @xmath69 the computational cost for calculating the contraction @xmath70 is @xmath71 , which is linear in @xmath0 . the matrix - vector product , or the linear mapping can also be efficiently represented by the tt format . the computational cost for computing a matrix - vector product in tt format is @xmath72 . 1 . suppose that a vector @xmath73 is in tt format and a matrix @xmath74 is in matrix tt format , and both are represented as kronecker products @xmath75 and @xmath76 then the matrix - vector product is represented in tt format @xmath77 where @xmath78 and @xmath79 + we can get the same expression in the case that @xmath80 and @xmath81 are represented as outer products . the linear mapping @xmath82 can also be represented by its entries written as products of matrices . suppose that @xmath80 and @xmath83 are in the tt formats @xmath84 and @xmath85 then the entries of the linear mapping @xmath86 is calculated by the contraction @xmath87 where the lateral slices @xmath88 of tt - cores @xmath89 are expressed as @xmath90 note that @xmath91 and @xmath92 are row and column vectors . 3 . we can further simplify the notation ( [ eqn : core_contraction_entry ] ) by considering the tt - core @xmath93 as an operator . let @xmath94 be a linear map defined by @xmath95 with each @xmath96th slice @xmath97 we can represent @xmath98 in a simplified form as follows . let @xmath99 be in matrix tt format and @xmath100 be in tt format . the linear mapping @xmath98 is represented by @xmath101 on the other hand , we recall that @xmath102 where @xmath103 is the frame matrix and @xmath104 . a large - scale matrix - vector multiplication reduces to a smaller matrix - vector multiplication as @xmath105 , where @xmath106 but we can not calculate @xmath107 by matrix - matrix multiplication for a large matrix @xmath108 . however , by using the representation ( [ eqn : linearmap_tensor_operation ] ) , we can show that @xmath107 can be calculated by recursive core contractions as follows . let @xmath109 be a linear map defined by @xmath110 for any @xmath111 . let @xmath107 be the matrix defined by . then @xmath112 for any @xmath111 . that is , the matrix - vector product @xmath113 , where @xmath114 , can be computed efficiently via the recursive core contractions in @xmath115 . figure [ fig : loc_lin_op ] illustrates the graph for the linear map @xmath116 . for each @xmath117 core tensor @xmath118 is connected to the core tensor @xmath119 , which is represented as @xmath120 in ( [ eqn : linearmap_local_operation ] ) . graph for the matrix - vector product @xmath121 represented by the linear mapping @xmath122 , where @xmath104 , width=264 ] the quadratic form @xmath123 for a symmetric and very large - scale matrix @xmath124 can be represented in tt format as follows . 1 . let @xmath108 and @xmath12 are represented as kronecker products @xmath75 and @xmath76 then the quadratic form is represented as the products @xmath125 where @xmath126 and @xmath127 for @xmath128 and @xmath129 for @xmath130 . the outer product representation leads to the same expression . recall that the @xmath2th core tensor of @xmath131 can be represented by the @xmath96th slices @xmath132 , @xmath133 . the quadratic form @xmath134 is computed by contraction of the hadamard product with the rank - one tensor as @xmath135 where @xmath136 we have @xmath137 and @xmath138 . 3 . instead , by using the core contraction defined in ( [ eqn : define_core_contraction ] ) , we can simplify the expression ( [ eqn : quadratic_mps_core ] ) by @xmath139 where @xmath140 is defined in the previous subsection . finally , we can express the quadratic form efficiently as @xmath141 on the other hand , from @xmath102 the quadratic form @xmath142 reduces to @xmath143 , where @xmath144 is a much smaller matrix than @xmath108 when tt - ranks @xmath145 and @xmath146 are moderate . since @xmath147 can not be calculated by matrix - matrix multiplication for a large matrix @xmath108 , we calculate it iteratively by recursive core contractions based on the distributed representation ( [ eqn : quadratic_tensor_represn ] ) as follows . let @xmath148 be a bilinear form defined by @xmath149 for any @xmath150 . let @xmath151 be the matrix defined by . then @xmath152 for any @xmath150 . that is , the bilinear form @xmath153 , where @xmath154 and @xmath114 , can be computed efficiently via the recursive core contractions in @xmath155 . figure [ fig : loc_quad_op ] illustrates the graph for the bilinear form @xmath156 . it is clear that each core tensor @xmath157 is connected to the core tensor @xmath119 , which is represented as @xmath158 in ( [ eqn : quadratic_local_represn ] ) . graph for the quadratic form @xmath159 represented by the bilinear form @xmath160 , where @xmath104 , width=283 ] in this paper , we proposed several new mathematical operations on tensors and developed novel representations of the tt formats . we generalized the standard matrix - based operations such as the kronecker product , hadamard product , and direct sum , and proposed tensor - based operations such as the self - contraction and core contraction . we have shown that the tensor - based operations are able to not only simplify traditional index notation for tt representations but also describe important basic operations which are very useful for computational algorithms . the self - contraction operator can be used for defining the tensor chain ( tc ) @xcite representations , and its properties should be more investigated in the future work . moreover , the definition of core contraction can also be generalized to any tensor network formats such as hierarchical tucker ( ht ) format @xcite . the partial contracted products of either the left or right core tensors are matricized and used as a building block of the frame matrices . we have shown that the suggested tensor operations can be used to prove the orthonormality of the frame matrices , which have been proved only by using index notation in the literature . the developed relationships may also play a key role in the alternating linear scheme ( als ) and modified alternating linear scheme ( mals ) algorithms @xcite for reducing the large - scale optimizations to iterative smaller scale problems . recent studies adjust the frame matrices in order to incorporate rank adaptivity and improve convergence for the als @xcite . in this work , we have derived the explicit representations of the localized linear map @xmath116 and bilinear form @xmath156 by the proposed tensor operations , which are important for tt - based iterative algorithms for breaking the curse - of - dimensionality @xcite . the global convergence of the iterative methods remains as a future work . in addition , it is important to keep the tt - ranks moderate for a feasible computational cost . a. cichocki , r. zdunek , a .- phan , and s. amari ( 2009 ) . _ nonnegative matrix and tensor factorizations : applications to exploratory multi - way data analysis and blind source separation _ , chichester : wiley . a. cichocki ( 2013 ) . era of big data processing : a new approach via tensor networks and tensor decompositions . invited talk at _ international workshop on sisa-2013 _ , nagoya , 1 oct . arxiv:1403.2048 . s. v. dolgov , b. n. khoromskij , i. v. oseledets , and d. v. savostyanov ( 2014 ) . computation of extreme eigenvalues in higher dimensions using block tensor train format . _ computer physics communications _ , 185(4 ) , 12071216 . s. v. dolgov and i. v. oseledets ( 2012 ) solution of linear systems and matrix inversion in the tt - format . _ siam j. sci . _ , 34 , a27182739 . s. v. dolgov and d. v. savostyanov ( 2013 ) . alternating minimal energy methods for linear systems in higher dimensions part i : spd systems . arxiv:1301.6068 . s. v. dolgov and d. v. savostyanov ( 2013 ) . alternating minimal energy methods for linear systems in higher dimensions . part ii : faster algorithm and application to nonsymmetric systems . arxiv:1304.1222 . a. falc and w. hackbusch ( 2012 ) . on minimal subspaces in tensor representations . _ , 12 , 765803 . l. grasedyck ( 2010 ) . hierarchical singular value decomposition of tensors . _ siam j. matrix anal . _ , 31(4 ) , 20292054 . s. holtz , t. rohwedder , and r. schneider ( 2012 ) . the alternating linear scheme for tensor optimization in the tensor train format . _ siam j. sci . _ , 34(2 ) , a683a713 . v. a. kazeev and b. n. khoromskij ( 2012 ) . low - rank explicit qtt representation of the laplace operator and its inverse . _ siam j. matrix analysis applications _ , 33(3 ) , 742758 . v. a. kazeev , b. n. khoromskij , and e. e. tyrtyshnikov ( 2013 ) . multilevel toepliz matrices generated by tensor - structured vectors and convolution with logarithmic complexity . _ siam j. sci . _ , 35(3 ) , a1511a1536 . t. g. kolda ( 2006 ) . multilinear operators for higher - order decompositions . technical report sand2006 - 2081 , sandia national laboratories . b. n. khoromskij and i. v. oseledets ( 2010 ) . dmrg+qtt approach to computation of the ground state for the molecular schrdinger operator . preprint 69 , mpi mis , leipzig . d. kressner , m. steinlechner , a. uschmajew ( 2013 ) . low - rank tensor methods with subspace correction for symmetric eigenvalue problems . mathicse technical report 40.2013 , epfl , lausanne . w. d. launey and j. seberry ( 1994 ) . the strong kronecker product . _ journal of combinatorial theory , series a _ , 66(2 ) , 192213 . doi:10.1016/0097 - 3165(94)90062 - 0 . i. v. oseledets and e. e. tyrtyshnikov ( 2009 ) . breaking the curse of dimensionality , or how to use svd in many dimensions . _ siam j. sci . _ , 31(5 ) , 37443759 .	we review and introduce new representations of tensor train decompositions for large - scale vectors , matrices , or low - order tensors . we provide extended definitions of mathematical multilinear operations such as kronecker , hadamard , and contracted products , with their properties for tensor calculus . then we introduce an effective low - rank tensor approximation technique called the tensor train ( tt ) format with a number of mathematical and graphical representations . we also provide a brief review of mathematical properties of the tt format as a low - rank approximation technique . with the aim of breaking the curse - of - dimensionality in large - scale numerical analysis , we describe basic operations on large - scale vectors and matrices in tt format . the suggested representations can be used for describing numerical methods based on the tt format for solving large - scale optimization problems such as the system of linear equations and eigenvalue problems .
water ice is believed to play many important roles in the planet formation theories . for example , ice enhances the surface density of solid material in the cold outer part of a protoplanetary disk , which promotes the formation of massive cores of gaseous planets ( e.g. , * ? ? ? * ) . thus the ice sublimation / condensation front called snowline , is considered to be the boundary of the forming regions of the terrestrial and jovian planets . snowline is also suggested as a possible forming site of the planetesimals @xcite . furthermore , icy planetesimals or comets may bring water to the earth ( e.g. , * ? ? ? recently , numerous water vapor emission lines have been detected in protoplanetary disks ( e.g. , * ? ? ? * ; * ? ? ? * ) , and the position of snow line is inferred from the modeling @xcite . on the other hand , observations of water ice distribution in the disk are limited at this moment . crystalline h@xmath1o ice emission features at 44 and 62@xmath0 m have been found for several herbig ae / be stars @xcite , but the limited angular resolution in far - infrared wavelengths hampers us to obtain ice distribution . while near - infrared ( nir ) water ice absorption toward edge - on disks is reported @xcite , the ice absorption is formed at somewhere through line of sight , thus it is still not straightforward to derive its radial distribution in the disk . @xcite proposed a new observational way to investigate the radial distribution of ice in face - on disks . they showed that ice absorption should also be imprinted in the light scattered by icy grains and that multi - wavelength imaging in nir wavebands , including h@xmath1o band at 3.1@xmath0 m , is a useful tool to constrain the ice distribution in the disk . @xcite applied this method to the circumstellar disk around a herbig fe star hd 142527 , and showed that the water ice grains present in a disk surface at a radial distance of 140 au . on the other hand , @xcite calculated the stability and distribution of water ice grains in the disk surface considering the photodesorption ( photosputtering ) process by uv irradiation . they showed that the water ice grains can be rapidly destroyed at the disk surface around a / b type stars due to uv photo desorption processes . although @xcite already detected the water ice grains in the disk surface around f - type star hd142527 , it would be interesting to observe the water ice grains in the disk surface around a / b type stars to check the prediction by @xcite . in this paper , we showed the observations of water ice grains in the disk surface around herbig be star hd100546 , and discuss the presence / stability of water ice grains in the disk surface . part of our data was already published in @xcite focusing the planet candidate @xcite direct imaging observations of the herbig be star hd 100546 using k band filter ( central wavelength @xmath2 = 2.20@xmath0 m , and width @xmath3 = 0.33@xmath0 m ) , h@xmath1o ice filter ( @xmath2 = 3.06@xmath0 m , @xmath3 = 0.15@xmath0 m ) and l band filter ( @xmath2 = 3.78@xmath0 m , @xmath3 = 0.70@xmath0 m ) were performed using the nici ( near infrared coronagraphic imager ; * ? ? ? * ) on the gemini south telescope on march 31 , 2012 . we fixed the instrument rotator during the observations of both object and psf ( point - spread function ) reference stars to fix the pupil and to obtain the stable psf patterns . a full width at half maximum ( fwhm ) of 0.10@xmath4 was achieved at all the wavelength using the instrument ao system . the central region close to the central star was saturated , however , the outer part ( r @xmath5 0.22@xmath4 ) is not saturated and can be used for disk observations . the total exposure times were 1672 s , 3192 s , and 2128 s for k , h@xmath1o ice , and l , respectively . as a psf reference star , we observed hr 4977 ( a0v ) just before / after hd 100546 . hd 105116 ( k band ) and bs4638 ( h@xmath1o ice and l band ) was observed as a photometric standard star . the mean flux density for bs 4638 at 3.06@xmath0 m was estimated by scaling the kurucz s stellar model atmosphere ( @xmath6 = 19500 k , @xmath7 = 3.95 , solar metallicity ) to match the flux density in k @xcite . the calculated flux density was 5.87 jy at 3.06@xmath0 m . observation parameters are summarized in table [ obssummary ] . the images were first processed using the iraf packages for dark subtraction , flat - fielding with sky flats , bad pixel correction , and sky subtraction . since the stellar halo was very bright , psf subtraction was required to investigate the faint structure near the central star . the reference psf was chosen to match the psf of hd 100546 for each frame with careful visual inspection to determine whether the circular bright halo of the central point source was well suppressed after psf subtraction . the reference psf was made by combining the adopted reference star images . the flux scaling of the reference psf was performed so that no region had negative intensity after the subtraction , in particular , just outer part of central radius ( r @xmath8 ) . the reference psf was shifted to match the central position and subtracted from each frame of hd 100546 . the each subtracted frame was rotated to match the north direction . then , the final psf subtracted image was made by combining the rotated object frames . in order to estimate the systematic uncertainty of the surface brightness due to the psf subtraction process , we changed the scaling of the psf before subtraction , and measured the acceptable range of the scaling factor . the systematic uncertainty of the surface brightness was measured to be 20% depending on the position . this systematic uncertainty was typically larger than the statistical error derived from the standard deviation of the best psf - subtracted object frames . the final psf subtracted images of hd100546 disk is shown in figure [ obsimages ] . in figure [ obsimages ] , an extended disk structure is detected in all three bands . especially in l band , a dark lane is seen in the south - west direction from the star , showing the typical inclined flaring disk morphology . the north - east side is facing to us , while the south - west scattered light beyond the dark lane can be the scattered light from the other side of the disk surface . this morphology is consistent with the previous studies @xcite . using these three color images , we extracted the scattered light spectra of different region of the protoplanetary disk . since the position angle ( pa ) of this disk major axis is 145@xmath9 @xcite , we set the 0.162@xmath4(9 pixel ) square region along with major and minor axis at a distance of 0.360@xmath4 , 0.522@xmath4 , 0.684@xmath4 , 0.846@xmath4 , and 1.008@xmath4 from the central star , which is shown in figure [ fig : extract ] . the extracted spectra of the each region are shown in figure [ fig : spectra ] . in almost all the regions , relatively shallow 3@xmath0 m absorption feature is present in their spectra likely due to water ice grains , indicating that the water ice grains present in the disk surface . the shallowness of this ice absorption feature can be due to the loss of ice grains at the disk surface . @xcite claimed that the water ice grains can be quickly destroyed at the disk surface around a / b type stars due to its strong uv photodesorption . to assess their prediction , quantitative comparison with the model prediction is required . since we are interested in the water ice distribution in the disk and it is necessary to quantify the absorption feature depth for comparison with the disk model , we will use the water ice absorption optical depth @xmath10 following the convention . note that the absorption feature in the scattered light spectra is not a pure absorption but rather an albedo effect ( see @xcite ) , however , the optical depth @xmath10 is so often used as an indicator of the depth of the feature , thus we use it for descriptive purposes . the optical depth is derived from the following formula as usual . where @xmath15 and @xmath16 are the observed surface brightness at k and l filters , respectively , while @xmath17,@xmath18 , and @xmath19 are the central wavelengths of k , h@xmath1o and l filters , respectively . as the ice absorption depth becomes deeper , the optical depth @xmath10 value becomes larger . we derived the @xmath10 at each extracted disk regions , and these values ranged from 0 to 2 . the @xmath10 map is shown in fig.[fig : h2otaumap ] . in this figure , the central region ( @xmath200.22 ) is masked due to the saturation problem and outer region ( [email protected] ) is also masked because of the low signal to noise ratio . in general , south - western(sw ) region shows relatively high @xmath10 value ( @xmath21 ) , which coincides with the dark - lane seen in l disk image . this can be qualitatively understood that some scattered light of this region may come from the backside of the disk , which suffers more extinction than that from the frontside of the disk . other than the sw region , a possible trend could be recognized that the inner region shows lower optical depth ( @xmath22 ) which might imply the decrease of the ice grains toward the central star as expected , although the patchy structure in the @xmath10 map hampers us to make a solid conclusion . since the @xmath10 map shows significant asymmetry along the disk minor axis ( sw - ne ) and scattered light intensity depends on the scattering geometry , discussion on the scattered light spectra along the disk minor axis is more complicated than that along the disk major axis ( se - nw ) . thus we focus on a comparison with the model along the disk major axis . the radial distributions of @xmath10 along with the disk major axis are summarized in table[measuresummary ] and shown in fig.[fig : h2otau ] . as already described , the error is dominated by systematic error , not the statistical one , and is estimated 20% of the surface brightness at each wavebands . to discuss whether the observed absorption feature is consistent with the model predictions , we derived the expected @xmath10 value based on the disk model calculations by @xcite who included the effect of photodesorption of water ice grains by uv photons from the central star . our calculation model consists of two parts : one is the disk structure calculation ( model part 1 ) in which the density and temperature distributions in the disk , as well as the snow line , are obtained , and the other is the radiative transfer calculation ( model part 2 ) that simulates observations . in the model part 1 , we obtain the location of the snow line , which is primarily determined by two reasons ; the thermal sublimation and the photodesorption of water ice particles . in order to evaluate the thermal sublimation of ice particles , we obtain the temperature and the gas density distributions in the disk as follows . the temperature in the disk is determined by the energy balance between heating and cooling . heating sources for the disk include the radiation from the central star illuminating the surface of the disk and the viscous heating due to the disk accretion . cooling process is the radiative transfer , which finally emits energy from the disk to outer space by means of radiation . the surface density distribution is provided as a model parameter , and the gas density distribution along the vertical direction with respect to the disk is determined so that the hydrostatic equilibrium is achieved . the temperature in the disk and the shape of the disk surface affect each other , because the angle between the direction of the light from the central star and the disk surface determines the radiative energy received by the disk surface , and the inclination of the disk surface is a function of the temperature distribution . thus , the temperature distribution and the gas density distribution , which is related to the disk surface , should be calculated consistently . using the temperature and the gas density , we can evaluate the vapor pressure and the saturated vapor pressure of water , which are used to determine the snow line . on the other hand , we calculate the radiative transfer of uv radiation from the central star to the disk and obtain the uv flux exposed to ice particles in the disk , which gives the photodesorption rate . finally , we can see if ice particles can be present stably with the vapor pressure ( contributing to the condensation ) , the saturated vapor pressure ( contributing to the sublimation ) , and the photodesorption rate , and we obtain the location of the snow line in the disk . for our calculations , we use physical parameters for hd 100546 given by @xcite , which are listed in table [ modelparameter ] . we adopt the mass fractions of silicate and water to the disk gas of 0.0043 and 0.0094 , respectively @xcite . we also assume that they form spherical dust particles , that ice particles contain pure ice and silicate grains are composed of only silicate , that radii of dust particles range from 0.025 @xmath0 m to 2.5 @xmath0 m , and that their size distributions follow the power - law with the index of -3.5 . absorption and scattering coefficients of dust particles are given by @xcite . to make a comparison with observation ( model part 2 ) , we carry out monte carlo radiative transfer calculations . deviations of obtained results seen in fig . [ fig : h2otau ] are caused by statistical errors intrinsically related to monte carlo simulations . in the calculations , we take into account the anisotropic scattering and polarization produced by dust particles . the scattering matrices of silicate and ice particles are obtained using the bhmie code by @xcite with the complex refractive indices by @xcite . the radial profiles of the @xmath10 of disk models with or without photodesorption along with the disk major axis are also shown in fig.[fig : h2otau ] . a model without photodesorption effect shows deeper water ice absorption feature ( larger @xmath10 ) than the observations , while models with photodesorption effect show relatively shallow absorption ( smaller @xmath10 ) . this is due to the destruction of water ice grains at the disk surface via photodesorption by strong irradiation of uv photons from the central star @xcite . although the observed @xmath10 values match with both disk models with / without photodesorption effect , the model with photo desorption effect seems slightly better match with the observations at least for nw region . it would be interesting to note that the water vapor is reported to be depleted in the disk atmosphere of herbig aebe stars by the observations of water vapor lines @xcite . a plausible explanation is due to photodissociation of water by uv photons in the disk atmosphere . although photodissociation effect is not included in the model of @xcite , they discussed that the photodesorption process is much important for the water ice stability , while photodissociation effect is crucial for water vapor destruction in the disk surface . in any cases , the uv photons seem to play important role on the survival of both water ice and gas in the disk surface around herbig ae / be stars . it is apparent that our data do not have a high signal - to - noise ratio enough to distinguish the models with / without photodesorption process . this is because the systematic error dominates over the total error . further observations with better photometric accuracy are strongly desired . since these observations shown here employed techniques similar to so - called lucky imaging technique , improvement of systematic error is principally limited . however , when we make use of the polarimetric differential imaging ( pdi ) and/or spectroscopy , the systematic error can be significantly reduced and it will change the situation dramatically . thus the l - band pdi and/or spectroscopy is promising for the advance of this observations . furthermore , other effects on the depth of water ice absorption , such as grain size , grain shape , grain structure , ice / rock ratio ( abundance ) , dust settling , turbulent mixing , and so on , should be investigated in the future theoretical studies to comprehensively understand the water ice distribution in the protoplanetary disks . we are grateful to all of the staff members of the subaru telescope , gemini telescope , and optical coatings japan for the production of narrow band filters . we also thank anonymous referee for their useful comments . we appreciate dr . tom hayward on his kind support during the observations and installation of our h@xmath1o ice filter to nici . mh was supported by jsps / mext kakenhi ( grant - in - aid for young scientists b : 21740141 , grant - in - aid for scientific research on innovative areas : 26108512 ) . mt is partly supported by the jsps fund ( no . 22000005 ) . fukagawa , m. , tamura , m. , itoh , y. , kudo , t. , imaeda , y. , oasa , y. , hayashi , s. s. , & hayashi , m. 2006 , , 636 , l153 honda , m. , et al . 2010 , , 718 , l199 hayashi , c. , nakazawa , k. , & nakagawa , y. 1985 , protostars and planets ii , 1100 brauer , f. , henning , t. , & dullemond , c. p. 2008 , , 487 , l1 morbidelli , a. , chambers , j. , lunine , j. i. , et al . 2000 , meteoritics and planetary science , 35 , 1309 carr , j. s. , & najita , j. r. 2008 , science , 319 , 1504 salyk , c. , pontoppidan , k. m. , blake , g. a. , et al . 2008 , , 676 , l49 malfait , k. , waelkens , c. , bouwman , j. , de koter , a. , & waters , l. b. f. m. 1999 , , 345 , 181 meeus , g. , waters , l. b. f. m. , bouwman , j. , et al . 2001 , , 365 , 476 pontoppidan , k. m. , dullemond , c. p. , van dishoeck , e. f. , et al . 2005 , , 622 , 463 terada , h. , tokunaga , a. t. , kobayashi , n. , et al . 2007 , , 667 , 303 inoue , a. k. , honda , m. , nakamoto , t. , & oka , a. 2008 , , 60 , 557 oka , a. , inoue , a. k. , nakamoto , t. , & honda , m. 2012 , , 747 , 138 honda , m. , inoue , a. k. , fukagawa , m. , et al . 2009 , , 690 , l110 chun , m. , toomey , d. , wahhaj , z. , et al . 2008 , , 7015 , 70151v allen , d. a. , & cragg , t. a. 1983 , , 203 , 777 carter , b. s. , & meadows , v. s. 1995 , , 276 , 734 van boekel , r. , min , m. , waters , l. b. f. m. , et al . 2005 , , 437 , 189 quanz , s. p. , schmid , h. m. , geissler , k. , et al . 2011 , , 738 , 23 mulders , g. d. , min , m. , dominik , c. , debes , j. h. , & schneider , g. 2013 , , 549 , aa112 mulders , g. d. , waters , l. b. f. m. , dominik , c. , et al . 2011 , , 531 , aa93 soubiran , c. , le campion , j .- f . , cayrel de strobel , g. , & caillo , a. 2010 , , 515 , aa111 currie , t. , muto , t. , kudo , t. , et al . 2014 , , 796 , ll30 miyake , k. , & nakagawa , y. 1993 , icarus , 106 , 20 bohren , c. f. , & huffman , d. r. 1983 , new york : wiley , 1983 , avenhaus , h. , quanz , s. p. , meyer , m. r. , et al . 2014 , , 790 , 56 fedele , d. , pascucci , i. , brittain , s. , et al . 2011 , , 732 , 106 zhang , k. , pontoppidan , k. m. , salyk , c. , & blake , g. a. 2013 , , 766 , 82 quanz , s. p. , amara , a. , meyer , m. r. , et al . 2013 , , 766 , ll1 ccccc hd100546 & k & 1672 & + hr4977 & k & 456 & psf reference + hd105116 & k & 152 & photometric reference + hd100546 & h@xmath1o ice & 3192 & + hr4977 & h@xmath1o ice & 2280 & psf reference + bs4638 & h@xmath1o ice & 152 & photometric reference + hd100546 & l & 2128 & + hr4977 & l & 532 & psf reference + bs4638 & l & 152 & photometric reference rrrrr + 104 & [email protected] & [email protected] & [email protected] & [email protected] + 87 & [email protected] & [email protected] & 65@xmath2313 & [email protected] + 70 & 60@xmath2312 & [email protected] & 111@xmath2322 & [email protected] + 54 & 126@xmath2325 & 71@xmath2314 & 210@xmath2342 & [email protected] + 37 & 304@xmath2361 & 223@xmath2345 & 485@xmath2397 & [email protected] + + 104 & [email protected] & [email protected] & [email protected] & [email protected] + 87 & [email protected] & [email protected] & 63@xmath2313 & 0.76@xmath23 0.39 + 70 & 52@xmath2310 & [email protected] & 109@xmath2322 & 0.74@xmath23 0.39 + 54 & 106@xmath2321 & 92@xmath2318 & 219@xmath2344 & 0.59@xmath23 0.39 + 37 & 250@xmath2350 & 237@xmath2347 & 506@xmath23100 & 0.50@xmath23 0.39 lc stellar effective temperature @xmath24 & 10500 k + stellar mass @xmath25 & 2.4 @xmath26 + stellar luminosity @xmath27 & 36 @xmath28 + stellar radius @xmath29 & 1.8 @xmath30 + distance d & 103 pc + disk inner radius @xmath31 & 0.5 au + disk outer radius @xmath32 & 500 . au + disk inclination & 45@xmath9 + surface density at 1au & 45 g/@xmath33 + surface density power - law index q & 1	we made near infrared multicolor imaging observations of a disk around herbig be star hd100546 using gemini / nici . k ( 2.2@xmath0 m ) , h@xmath1o ice ( 3.06@xmath0 m ) , and l(3.8@xmath0 m ) disk images were obtained and we found the 3.1@xmath0 m absorption feature in the scattered light spectrum , likely due to water ice grains at the disk surface . we compared the observed depth of the ice absorption feature with the disk model based on @xcite including water ice photodesorption effect by stellar uv photons . the observed absorption depth can be explained by the both disk models with / without photodesorption effect within the measurement accuracy , but slightly favors the model with photodesorption effects , implying that the uv photons play an important role on the survival / destruction of ice grains at the herbig ae / be disk surface . further improvement on the accuracy of the observations of the water ice absorption depth is needed to constrain the disk models .
ultracold atoms provide an excellent forum to study complex quantum mechanical behavior . an example is the superfluid to mott insulator transition @xcite , where experimental efforts have imaged this as a quantum phase transition at the single atom level @xcite . a quantum phase transition is a fundamental change in the ground state as a parameter is altered , in the superfluid - mott insulator example the parameter is lattice depth . near this quantum phase transition the temperature dependence of superfluidity and quantum criticality has been studied @xcite . another system in ultracold atoms , which exhibits a quantum phase transition , is the 2 component bose einstein condensate ( bec ) , where by tuning the interactions between the components the gas can change from a miscible to an immiscible phase , as has been experimentally demonstrated in rb @xcite . theoretical studies showed that if @xmath0 is greater ( less ) than zero , the gas is miscible ( immiscible ) @xcite . here @xmath1 is the coupling strength in the mean field treatment between the @xmath2 and @xmath3 component . interestingly , 2 component becs have been used to create vortices @xcite and study non - equilibrium dynamics @xcite . in this work , we look at the character of the excitations across the quantum phase transition in the 2 component bec system . for a miscible system , the modes are collective in nature , and the condensates move either in - phase or out - of - phase with each other . for the immiscible system , the excitations are either collective or interface excitations . this work offers a new perspective into the nature the miscible - immiscible transition with characterization of the excitations . there has been previous work on excitation spectra for trapped 2 component bec @xcite . that work studied symmetry breaking as a function of particle number and found that a mode goes to zero when the system becomes immiscible , but they did not study the nature of the quasiparticles . ref . @xcite studied the ground state and characterized the mode that goes soft ( energy goes to zero ) , here we extend the analysis to many low - lying modes of the system . to obtain the excitation spectrum , we solve the bogoliubov de gennes equations @xcite for a trapped gas with contact interactions . first , we must solve the gross pitaevskii equation for the 2 condensates ( @xmath4 ) : @xmath5 @xmath6 is the kinetic energy and trapping potential and @xmath7 is the chemical potential for the @xmath2 component . we normalize @xmath4 so that @xmath8=1 . now we consider the ground state and its excitations to be of the form : @xmath9 . substituting this into the time dependent version of eq . ( [ gpe ] ) and collecting powers of @xmath10 and linear terms in @xmath11 , we find the excitations are given by : @xmath12 we have assumed @xmath4 is real , and that the different components have equal number and mass . the second term contains both the exchange and anomalous term , which couples @xmath13 to @xmath14 and @xmath13 to @xmath15 , respectively . these terms would be non - local if the interaction was finite range . these excitations are normalized in the standard way : @xmath16-@xmath17 @xcite . to perform this work , we focus on the quasi2d case where high resolution experimental imaging is possible @xcite . for clarity , we will focus on two examples : one is miscible and the other immiscible , both far away from the transition so the character of the excitations is clear . we consider @xmath18100 and @xmath19=1.01 g ; for the miscible example , we pick @xmath20=0.5 g and an immiscible example , we pick @xmath20=2 g , where @xmath21 is the strength of the contact interaction , @xmath22 is the 3d s - wave the scattering length ( @xmath23 ) , @xmath24 is the axial harmonic oscillator length and @xmath25 is the trapping frequency in the tightly confined direction . we only consider @xmath26 and equal masses . we rescale the equations into trap units , so the energy scale is @xmath27 and the length scale is @xmath28 where @xmath29 is the trapping frequency in the x - y plane . we can loosely relate this to experiments , for g=100 if we pick @xmath30=1000 , @xmath25/@xmath29=100 , and @xmath22=100 @xmath31 , then this example corresponds to radial trapping frequencies , of @xmath3238 hz for k and @xmath3211 hz for cs . it is worth mentioning that the chemical potentials for each component are about equal ( @xmath33 ) . more importantly , for the immiscible system , @xmath34 is @xmath35 and this gives a healing length of @xmath36 ( for the miscible system @xmath37 ) . further details of how we solve these equations ( [ gpe],[bdg ] ) appear in ref . @xcite . only slightly less than @xmath38 , and in ( b ) we show the quantum gas just after quantum phase transition when the ground state breaks rotational symmetry and @xmath20 only slightly greater than @xmath38 . for ( c ) we have shown the standard example of an immiscible gas with @xmath39 . the bec labeled 1 ( 2 ) is shown as blue ( red ) . ] in fig . [ ground ] we see the ground state changes character as we vary @xmath20 . the component with the smaller @xmath40 is more dense in the middle of the trap . we define @xmath38 as the value of @xmath20 when the ground state changes character to a broken symmetry state which begins the immiscible regime . when @xmath41 , the ground state of the system has azimuthal symmetry and the 2 becs overlap , ( a ) . however as @xmath20 is increased to @xmath38 , the ground state suddenly changes , and the azimuthal symmetry is broken , see ( b ) . as @xmath20 is further increased , the 2 becs separate further and decrease their overlap as it becomes energetically costly , ( c ) . a similar evolution of the ground state as a function of @xmath20 was reported in ref . @xcite . it is challenging to find the ground state for all @xmath20 . to do so , we use the conjugate gradient method . we have found the best initial guess is one with a slightly broken symmetry and poor overlap with the final group state . we seed the noise so that the interface would be along the y axis . if the overlap between the initial guess and the ground state is too large then it is easy to get stuck in a local energy minimum . when the conjugate gradient method fails to find the true ground state , the solutions to the bogoliubov de gennes equations have complex eigenvalues when we vary @xmath20 , the previous solution is thrown out . in this way we reliably find the excitation spectrum of the 2 component bec with only real eigenvalues . is at 1.04 g for trapped example with @xmath42=100 . note the presence of strong avoided crossings and broken degeneracies after rotational symmetry has been broken , @xmath43 . for the miscible system as @xmath42 is increased , the out - of - phase collective excitations dramatically lower in energy . some of the excitations shown in fig . [ quasi ] ( [ quasi2 ] ) are labeled on the left ( right ) of this figure . ] in fig . [ energytrans ] we show bogoliubov excitation energies ( @xmath44 ) as a function of @xmath20 . this shows the transition from miscible to immiscible at @xmath45 , where a mode goes soft . homogeneous theory predicts this transition at @xmath46 @xcite . the discrepancy is explained by the trap and finite size of the gas @xcite . in fact , if we were to increase g ( keep @xmath47 fixed ) ; @xmath48 decreases toward 1 . for example if g=400 ( @xmath49 ) then @xmath50 . a recent study explored how the trapping and the kinetic energy contributions impact the criteria for immisciblity @xcite . our findings are consistent with their results . to further understand fig . [ energytrans ] , we start with @xmath41 where the systems is miscible . the quasi - particles are readily classified based on their azimuthal symmetry and the relative motion of the 2 condensates . as @xmath20 increases towards @xmath38 , many modes decrease in energy . then at @xmath38 two modes go soft or their energies go to zero . for @xmath43 , many degeneracies are broken , and there are many avoided crossing as @xmath20 is further increased . this is where the excitations mix and change character . for the miscible side of spectrum , there are energy crossings , but they are protected symmetry and do not couple . to further understand this transition , we look at the mode which goes soft at @xmath38 . with @xmath51 . energy is shown for each quasi - particle in trap units . ] in fig . [ quasi ] we show density perturbations associated with 2 low - energy excitations for the miscible system ( far left of fig . [ energytrans ] with @xmath51 ) . density perturbations from bogoliubov de gennes theory ( t=0 ) are given by @xmath52 for the mode @xmath53 . in fig . [ quasi ] we show both ( a ) the out - of - phase and ( b ) the in - phase slosh modes . the condensates for this example are co - spatial and nearly identical . they look similar to those in fig . [ ground ] ( a ) . in this figure , we separated the two components for clarity . the density moves from the regions define by the dashed lines to the regimes defined by the solid lines . the color of the perturbations matches the color for the associated condensate . we have drawn arrows to illustrate the motion of the density perturbations . the contours are shown for 0.25 , 0.5 and 0.75 of the maximum value of the perturbation , and condensate density is shown in the background . the energy of the mode is reported on the figure in trap units . in fig . [ quasi ] ( a ) we show a slosh mode , but the motions of the 2 becs are out - of - phase with each other . the solid lines coincide with the dashed lines for the other condensate s motion . for example , the blue condensate , ( i ) , sloshes from @xmath54 to @xmath55 while the red condensate , ( ii ) , sloshes from @xmath55 to @xmath54 . there is no center - of - mass motion in this case . in contrast , ( b ) shows a mode with center - of - mass motion which is a kohn mode of energy 1@xmath27 . the motion of each condensate coincides with the other ; both the blue and red condensate slosh from @xmath55 to @xmath54 in phase with each other . it is important to note that the out - of - phase slosh mode has the lowest energy , and goes soft at the quantum phase transition . for @xmath41 , modes with @xmath56 , where @xmath57 is the azimuthal quantum number , have a degenerate twin . in our case with real quasi - particle modes , degenerate modes are related by a rotation of @xmath58 . for example in [ quasi ] ( a ) and ( b ) there are degenerate twins are just rotated by @xmath59 . referring back to the energy spectra in fig . [ energytrans ] , we see that as @xmath20 increases towards @xmath38 the energy of the mode in fig [ quasi ] ( a ) decreases ( while ( b ) stays at @xmath60 ) . more generally , all out - of - phase modes significantly lower in energy as @xmath42 is increased to @xmath38 . in fact , the out - of - phase modes with @xmath61 , @xmath62 , @xmath63 , and @xmath64 are all below @xmath60 at @xmath38 . then at @xmath38 the energy of the out - of - phase slosh goes to zero and one mode stays zero for @xmath43 . the other mode ( rotated by @xmath59 ) shoots up in energy as @xmath20 is further increased beyond @xmath38 . the ground state spontaneously breaks rotational symmetry - which the hamiltonian has - and this leads to an extra zero energy goldstone mode in the excitations spectrum @xcite . there are already two goldstone modes associated with broken phase symmetry of each condensate , and they are : @xmath65 and @xmath66 . if one looks more closely at the third mode with @xmath67 , one finds it is a rotation of the interface ( @xmath68 ) . this extra zero energy mode has already been observed in the 2 component becs @xcite . goldstone modes have been discussed in more detail for spinor becs @xcite , and ref . @xcite found similar behavior for a goldstone mode in an attractive condensate with a bogoliubov de gennes treatment . as we have said for the miscible system , the classification of the modes is simple : we use azimuthal symmetry and relative motion of the two condensates . but for the immiscible system , there is no rotational symmetry , so the characterization of the excitations must be different . to study this in more detail , we look at several quasi - particles . . the x and y axes are in trap units . ( a , b ) are slosh modes , ( c , d ) are quadrupole modes ( e , f ) are breathing modes , and ( g , h ) are interface modes . the energy is reported in trap units and for this example @xmath69 . ] in fig . [ quasi2 ] we show the density perturbations associated with quasi - particle excitations for the immiscible system ( @xmath70 , @xmath71 , @xmath72 , and g=100 ) . there are two types of quasi - particle modes : first , bulk excitations which look like those from a standard condensate , and second , interface excitations where the excitations are localized to the interface between the 2 condensates . since there is no azimuthal symmetry , to classify the bulk modes we need to access how their motion is oriented relative to the interface . in fig . [ quasi2 ] , to depict the motion of the density perturbations , we show arrows in a few examples . the density moves from the dashed regions to the solid regions . we also show the energy of the excitation in trap units . the contours are shown for 0.25 , 0.5 and 0.75 of the maximum value of the perturbation and condensate density shown in the background . the collective modes look like those in a standard bec , however the two becs now act collectively to retain the excitation character . first , we look at the slosh modes of the systems . they are shown in fig . [ quasi2 ] ( a , b ) . in ( a ) the slosh mode with the center - of - mass displacement parallel to the interface ( kohn mode ) is shown and in ( b ) a slosh mode and center - of - mass displacement is perpendicular to interface ( also a kohn mode ) is shown . the arrows show that in ( a ) both the blue and red condensate sloshes from @xmath55 to @xmath54 . for example ( b ) both the blue and red condensate sloshes from @xmath73 to @xmath74 . next we show the quadrupole modes in fig . [ quasi2 ] ( c , d ) . ( c ) shows a quadrupole mode with a nodal line along interface , and ( d ) shows a quadrupole mode where the density increases at interface . these excitations are typical of a single component bec where the excitations are related by a @xmath58 rotation . but in this case , the 2 becs collude to make the excitation . additionally , these two excitations are very similar in energy . we show two breathing modes in fig . [ quasi2 ] ( e , f ) . they can be classified as in - phase and out - of - phase motion of the 2 condensates . in ( e ) we show an in - phase breathing mode , where both becs inhale at once , or they both move into or out of the center of the trap in unison . in ( f ) we show an out - of - phase breathing mode , where one condensate inhales and the other exhales . the energies of these modes are notably different : 2 and 2.73 @xmath27 . there is another class of excitation in the immiscible 2 component bec : interface excitations . two examples are shown in fig . [ quasi2 ] ( g , h ) . these excitations are localized along the interface , and in general they are out - of - phase excitations , i.e. the density of one moves to where the other is leaving . note these are low energy excitations , in fact ( g ) is the lowest energy excitations , for the system at 0.58 @xmath27 and ( h ) is only slightly higher than the two kohn modes at 1.08 @xmath27 . if @xmath42 is increased the mode in ( h ) will decrease below 1@xmath27 . so if the chemical potential is increased , then the interface modes become lower in energy . to illustrate this , we look at how the excitation energies change as a function of @xmath42 while keeping the ratio of the interactions fixed ( @xmath75 ) . this is shown in fig [ energyn ] ( a ) . for reference , on the far right where @xmath42=600 and @xmath76 , we have marked the interface modes with red @xmath77 s , there are 16 interface modes with energy under 3.5@xmath27 . in contrast , on the far left where @xmath51 , we have marked the interface modes with red @xmath78 s and there are only 7 modes under 3.5@xmath27 . as one moves up in excitation energy , each new excitation simply adds another bend to the interface . one more important point of fig . [ energyn ] ( a ) , is that as @xmath42 increases the energies of interface modes decrease . ) . the quasiparticles from fig . [ quasi2 ] are on the far left side of spectrum . the interface excitations are marked by @xmath77 or @xmath78 on either side of the figure . the energies of the quasiparticle shown in fig . [ quasi2 ] ( g , h ) are labeled . in ( b ) and ( c ) the density perturbations for the interface excitations when @xmath76 or @xmath42=600 , they are labeled on the right side of ( a ) . their energy is reported in trap units.,title="fig : " ] + ) . the quasiparticles from fig . [ quasi2 ] are on the far left side of spectrum . the interface excitations are marked by @xmath77 or @xmath78 on either side of the figure . the energies of the quasiparticle shown in fig . [ quasi2 ] ( g , h ) are labeled . in ( b ) and ( c ) the density perturbations for the interface excitations when @xmath76 or @xmath42=600 , they are labeled on the right side of ( a ) . their energy is reported in trap units.,title="fig : " ] ) . the quasiparticles from fig . [ quasi2 ] are on the far left side of spectrum . the interface excitations are marked by @xmath77 or @xmath78 on either side of the figure . the energies of the quasiparticle shown in fig . [ quasi2 ] ( g , h ) are labeled . in ( b ) and ( c ) the density perturbations for the interface excitations when @xmath76 or @xmath42=600 , they are labeled on the right side of ( a ) . their energy is reported in trap units.,title="fig : " ] fig . [ energyn ] ( b ) and ( c ) show two examples of the higher energy interface modes . first , fig . [ energyn ] ( b ) is a mode with 6 bends in the interface . an interesting point about ( c ) is that it is a hybrid mode , it also has some density perturbations near the edge of the gas , away from the interface . the mode shown in ( c ) is in a region where several modes are crossing and their character is changing . if we further increase @xmath42 , the interface mode lowers in energy and loses it collective nature and looks more like mode in ( b ) . related excitations have been studied in non - equilibrium simulations of 2 component becs where rayleigh - taylor instabilities have been predicted @xcite . in these studies , the value of @xmath79 is changed and this drives one bec into the other , the interface then becomes unstable and a rayleigh - taylor instability forms . in conclusion , we have characterized the excitations of a 2 component bec within the bogoliubov de gennes framework . we found that as @xmath20 is increased from a miscible regime to an immiscible regime , fig . [ energytrans ] , generally , all of the out - of - phase excitations lower in energy . the energy of the out - of - phase slosh mode goes to zero , fig . [ quasi ] ( a ) . this mode becomes new goldstone modes when the rotational symmetry is spontaneously broken in the immiscible system . we looked at the excitations of the immiscible systems when @xmath80 in fig . [ quasi2 ] . we found that there are bulk modes which look similar to the excitations of a single trapped bec . there are also excitations localized at the boundary between the condensates . one of these interface modes is lowest energy mode when the becs are immiscible , and there are many other low energy interface modes , see the red @xmath78 s or x s in fig . [ energyn ] ( a ) . furthermore , if one goes to a more strongly interacting bec regime ( while still immiscible ) , the interface modes lower in energy . future work will seek to understand the relationship between bogoliubov excitations across the miscible - immiscible transition and critical phenomena . the effect of temperature on this system will be studied within the hartree fock bogoliubov framework , where bogoliubov excitations are thermally occupied . additionally , we will look at how this collective excitation changes with non - local dipolar interactions @xcite . the author gratefully acknowledges support through a ldrd ecr grant , lanl which is operated by lans , llc for the nnsa of the u.s . doe under contract no . de - ac52 - 06na25396 . this research was supported in part by the national science foundation under grant no . nsf phy11 - 25915 . the author also gratefully acknowledges conversations with e. timmermans , l. a. collins , and r. m. wilson . m. greiner _ et al . _ , nature * 415 * , 39 ( 2002 ) . w. s. bakr _ nature * 462 * , 74 ( 2009 ) . s. trotzky _ et al . _ , nat * 6 * , 998 ( 2010 ) . _ , science . * 335 * , 1070 ( 2012 ) . s. b. papp , j. m. pino , and c. e. wieman , phys . * 101 * , 040402 ( 2008 ) . h. pu and n. bigelow , phys . lett . * 80 * , 1130 ( 1998 ) ; _ ibid . _ * 80 * , 1134 ( 1998 ) . e. timmermans , phys . lett . * 81 * , 5718 ( 1998 ) . m. r. matthews _ et al . _ , lett . * 83 * , 2498 ( 1999 ) . k. m. mertes _ et al . _ , lett . * 99 * , 190402 ( 2007 ) . r. p. anderson , c. ticknor , a. i. sidorov , and b. v. hall phys . a * 80 * , 023603 ( 2009 ) . e. nicklas _ et al . * 107 * , 193001 ( 2011 ) . d. gordon and c. m. savage , phys . a * 58 * 1440 ( 1998 ) . j. g. kim and e. k lee , phys . e * 65 * , 066201 ( 2002 ) . a. l. fetter and j. d. waleck , _ quantum theory of many - particle systems _ , ( dover , new york 2003 ) . . hung , x. zhang , n. gemelke , and c. chin , nature * 470 * , 236 ( 2011 ) . c. ticknor , phys . a * 85 * , 033629 ( 2012 ) ; phys . a * 86 * , 053602 ( 2012 ) . r. navarro , r. carretero - gonzlez , and p. g. kevrekidis , phys . a * 80 * , 023613 ( 2009 ) . l. wen _ et al . _ phys . rev . a * 85 * , 043602 ( 2012 ) . for a review see : t. brauner , symmetry * 2 * , 609 ( 2010 ) . t .- l . ho and v. b. shenoy , phys . * 77 * , 3276 ( 1996 ) . s. uchino , m. kobayashi , and m. udea , phys . a * 81 * , 063623 ( 2010 ) ; s. uchino , m. kobayashi , m. nitta , and m. ueda phys . rev . lett . * 105 * , 230406 ( 2010 ) . r. kanamoto , h. saito , and m. ueda , phys . rev . lett . 94 , 090404 ( 2005 ) . k. sasaki , n. suzuki , d. akamatsu , and h. saito , phys . a * 80 * , 063611 ( 2009 ) ; n. suzuki , h. takeuchi , k. kasamatsu , m. tsubota , and h. saito , phys . a * 82 * , 063604 ( 2010 ) ; h. takeuchi , n. suzuki , k. kasamatsu , h. saito , and m. tsubota , phys . b * 81 * , 094517 ( 2010 ) ; t. kadokura , t. aioi , k. sasaki , t. kishimoto , and h. saito , phys . rev . a * 85 * , 013602 ( 2012 ) . s. gautam and d. angom , phys . a * 81 * , 053616 ( 2010 ) . a. bezett , v. bychkov , e. lundh , d. kobyakov , and m. marklund , phys . a * 82 * , 043608 ( 2010 ) ; d. kobyakov , v. bychkov , e. lundh , a. bezett , and m. marklund , phys . a * 86 * , 023614 ( 2012 ) . r. m. wilson , c. ticknor , j. l. bohn , and eddy timmermans , phys . a * 86 * , 033606 ( 2012 ) .	we present analysis of the excitation spectrum for a 2 component quasi2d bose einstein condensate . we study how excitations change character across the miscible to immiscible phase transition . we find that the bulk excitations are typical of a single - component bec with the addition of interface bending excitations . we study how these excitations change as a function of the interaction strength .
this is an opening talk on diffraction at dis97 and we start with definitions . the longitudinal ( l ) and the two transverse ( t ) polarizations of the exchanged photon define the four components @xmath1 ( @xmath2 and @xmath3 ) in the expansion q^2xd^(5)(epepx ) dq^2dxdm^2dp_^2d = _ em \{(1-y+12y^2)d_t^d(3)+ ( 1-y)d_l^(3 ) . + ( 1-y)2d_tt^d(3)+(2-y ) d_lt^d(3)}/dm^2dp_^2d[eq:1.1 ] where @xmath4 is the diffractive mass , @xmath5 is the @xmath6 momentum transfer , and @xmath7 is the angle between the @xmath8 and @xmath6 planes . in the charged current ( cc ) ddis there are also the c- and p - odd , @xmath9 and @xmath10 , terms @xcite . each @xmath11 defines a set of * dimensionless * diffractive structure functions @xmath12 , = _ tot^pp16f_i^d(5)(,p_^2 , x_,,q^2 ) , [ eq:1.2 ] @xmath13 and @xmath14 the so - defined @xmath15 are smooth functions of @xmath16 , in contrast to the zeus / h1 definition of @xmath17 @xcite which blows up at @xmath18 due to the extra factor @xmath19 . here @xmath20 and @xmath21 are the ddis variables . the microscopic qcd mechanism of ddis , as developed in 1991 by nikolaev and zakharov @xcite , is a grazing , quasielastic scattering of multiparton fock states of the @xmath22 on the proton . ddis is best described viewing @xmath22 as a system of color dipoles spanning between quarks , antiquarks and gluons @xcite , interaction of which with the target proton is described by the color dipole cross section @xmath23 @xcite . here @xmath24 is the gluon structure function of the target nucleon and one sums the leading log@xmath25 pqcd diagrams for the color singlet two - gluon exchange . systematic expansion of ddis in excitations of the @xmath26 and higher fock states of the photon has been developed in @xcite . the technique @xcite has been a basis of all the subsequent pqcd works on ddis @xcite . the buchmuller et al . model @xcite is essentially identical to the color dipole picture ; their function @xmath27 is an exact counterpart of our @xmath28 evaluated in the pqcd born approximation . bjorken s aligned jet model @xcite is an integral part of the color dipole approach . for the 1996 status report at dis96 see @xcite . this extended presentation reviews also the new results reported at dis97 . vector meson production @xmath29 is an exclusive ddis . for the shrinkage of @xmath22 the typical dipole size ( the scanning radius ) contributing to the production amplitude equals @xmath30 and one finds @xcite : _ ^ 2(x_,r_s ) g^2(x_,q_l^2)q^2/(q^2+m_v^2)^4 . [ eq:2.1 ] for the factor 6 in @xmath31 the relevant pqcd scale , @xmath32 , is rather small : @xmath33 - 0.2 and @xmath34 - 0.15 @xcite . ( [ eq:2.1 ] ) is a basis of the very successful pqcd phenomenology of vector meson production ( @xcite and references therein ) . the exponent @xmath35 in @xmath36 , rises with @xmath37 due to the rise of the pqcd scales @xmath38 , in good agreement with the experiment . ddis into continuum at @xmath39 0.10.2 is dominated by the @xmath40 excitation . if @xmath41 is the transverse momentum of the @xmath42 jet w.r.t . the photon , then @xcite ~(1 - 2(k^2+m_f^2 m^2)g^2(x_,q_t^2 = m_f^2+k^2 ( 1- ) ) , + df_l^d(3 ) dk^2 ~1q^2g^2(x_,q_l^2 = m_f^2+k^2 ( 1- ) ) . [ eq:2.2 ] these results have been re - derived by several groups @xcite . because @xmath43 , we predict strong enhancement of large-@xmath44 jets . furthermore , @xmath45 is dominated by hard jets and scaling violations almost entirely compensate the higher twist behavior of @xmath46 @xcite which , for @xmath47 , completely takes over the leading twist transverse structure function @xcite , @xmath48 . the applicability of pqcd improves for large @xmath49 . one often fits @xmath50 . the rise of our fixed-@xmath4 pqcd scale @xmath51 predicts the rise of @xmath52 with @xmath37 . furthermore , at @xmath53 one enters the @xmath54 dominated region of @xmath55 and @xmath52 must become as large as for vector meson production , which is indeed seen in the fixed-@xmath4 data from zeus @xcite . this is one of the manifestations of the duality between the continuum and vector meson production derived in @xcite . for fixed @xmath49 , the exponent @xmath52 must not depend on @xmath37 production of @xmath40 dijets in ddis at @xmath56 , and/or by real photons , is especially interesting . in 1994 we showed that in this case the transverse momentum @xmath44 of jets probes directly the transverse momentum of gluons in the pomeron @xcite and @xmath57 . ddis at @xmath58 0.10.2 comes from excitation of the @xmath59 and higher fock states of the photon . it is dis off the pomeron sea with the typical sea structure function , @xmath60 which obeys an approximate triple - pomeron factorization conjectured in @xcite and derived in @xcite , f_2^d(4)(p_^2=0,x_,,q^2 ) f_2p(x_,q_sea^2 ) , [ eq:2.4 ] which is similar to the triple - pomeron limit of soft diffraction @xmath61 @xmath62 here @xmath63 0.2-@xmath64 is a propagation radius for perturbative gluons . in @xcite we showed that although @xmath65 0.15 - 0.2 gev@xmath66 is the dimensional quantity , it does not change much from @xmath67 to dis ; parametrically @xmath68 . the normalization of the pomeron sea structure function * predicted * in @xcite is in good agreement with the experiment , see @xcite and fig . 1 . cc ddis has several peculiarities @xcite : first , both the @xmath69 and @xmath70 receive extra contributions , and @xmath70 is anomalously large at small @xmath37 because of the nonconservation of weak currents , for the similar findings in standard dis see @xcite . second , in the large-@xmath49 cc ddis with @xmath71 beams one excites the @xmath72 pairs and the pqcd scale @xmath73 is different for ddis with leading charm , @xmath74 and leading ( anti)strangeness , @xmath75 . third , the @xmath76 distribution of the charm changes from the forward to backward hemispheres in the diffractive system . the interpretation of @xmath40 excitation as dis off the intrinsic @xmath40 of the pomeron must be taken with the grain of salt , because for the above variety of pqcd scales ( @xmath77 ) , the flavor , valence , sea and glue composition of the pomeron changes with @xmath78 @xcite . furthermore , because the pqcd scales @xmath79 depend on @xmath49 , and the @xmath78 and @xmath49 dependences are inextricably entangled , it is crystal clear that the often postulated @xcite , but never proven , diffractive and/or ingelman - schlein factorization , @xmath80 , with the process independent flux of pomerons in the proton @xmath81 and the @xmath78 independent structure function @xmath82 * does not exist in qcd . * besides this diffractive factorization breaking , we predicted an unprecedented dominance of the higher twist @xmath45 at @xmath47 , which makes the dglap evolution of @xmath83 completely invalid at large @xmath49 . still further , upon the @xmath76 integration , the factor @xmath84 $ ] in ( [ eq:2.1 ] ) transforms into the factor @xmath85 $ ] in @xmath86 @xcite . such an abnormal enhancement of a higher twist , @xmath87 , is unprecedented in dis . this higher twist is especially large at @xmath88 and is a further strong objection to the blind dglap evolution of @xmath86 . this higher twist is a likely explanation @xcite of the h1 finding @xcite of the rise of @xmath17 with @xmath37 even for large @xmath49 , for the related discussion see also @xcite . in the cc ddis there are still more higher twists which are due to the nonconservation of charged currents @xcite . the non - negotiable prediction from pqcd is a substantial intrinsic charm in the pomeron @xcite , the abundance of which for @xmath89 is predicted to rise from @xmath903% at @xmath91 to @xmath92 at @xmath93 . we predicted the charm abundance @xmath94 at @xmath56 from excitation of the @xmath95 and higher states . this intrinsic charm explains perfectly the diffractive charm data reported at dis97 @xcite without invoking unnatural hard glue in the pomeron . the above findings did not make their way yet into the widely used monte carlo codes for ddis , which are all still based on the discredited diffractive factorization and the dglap evolution of the large-@xmath49 diffractive structure function . as we have seen above , the transverse momentum structure of final states in ddis is quite different from that in the standard dis . an analysis @xcite of higher orders in ddis has shown that the structure of real and virtual radiative corrections in ddis and standard dis is different , and one needs more work on final state radiation before turning on hadron shower codes which were developed for , and fine tuned to , the standard dis final states . in conclusion , all the dglap analysis and the factorization model - based conclusions on the hard glue in the pomeron are suspect . one badly needs monte - carlo implementation of the color dipole picture with due incorporation of the above host of pqcd scales , which is not available at the moment . the situation changes at @xmath96 , where @xmath97 has been proven to satisfy the conventional dglap evolution in @xmath98 ( at least for a fixed @xmath78 ) @xcite . the obvious conclusions are the rise of @xmath99 as @xmath49 decreases and/or @xmath37 rises and ( still approximate ) applicability of the conventional parton model for hard scattering off the pomeron . incidentally , via unitarity the diffractive higher twist also implies substantial higher twist effects in the proton structure function at @xmath100 , which hitherto has been ignored in the qcd analysis of the hera data . the real reason behind the plethora of pqcd scales and effective intercepts @xmath52 which vary with @xmath101 and flavor , is that the qcd pomeron described by the running bfkl equation is a sequence of * moving * poles with intercepts @xmath102 @xcite . here @xmath103 labels the eigenfunctions @xmath104 of the running bfkl equation for the dipole cross section and equals the number of nodes in @xmath104 , for more details see zoller s talk at dis97 @xcite . the onset of the dominance of the rightmost pole with @xmath105 is elusive because of the substantial contribution from the subleading poles in the range of @xmath106 accessible at hera @xcite . besides the subleading pomeron poles , ddis at a moderately small @xmath78 picks secondary reggeon exchanges @xmath107 . the @xmath108 exchange in the @xmath109 scattering originates from dis off the valence quarks and is represented by color singlet @xmath40 exchange in the @xmath110-channel . because of the kinematical relation , @xmath111 , the @xmath108-reggeon exchange is most prominent in ddis at small @xmath49 and lowers the exponent @xmath112 in the fit @xmath113 , which is evident in the discussion of the proton spectra in @xcite and has indeed been observed by the h1 @xcite . eventually , one must be able to relate the @xmath108-exchange in ddis to the valence structure function of the proton and to calculate the structure function of the reggeon directly from pqcd @xcite . the existing triple - regge analyses @xcite rather use the model estimates for the flux of reggeons from hadronic diffraction and take for the reggeon the pion structure function . the major issue is the reggeon - pomeron interference . if one treats the pomeron and reggeon as `` orthogonal hadronic states '' , then the interference must be weak @xcite . to the contrary , the pqcd suggests @xcite a strong constructive reggeon - pomeron interference . the strong interference scenario is preferred by the h1 data shown in fig . 2 . the @xmath114 dependence of diffractive cross section is a new observable accessible with the advent of the lps era . the standard parameterization is @xmath115 . the so - defined diffraction slope @xmath116 is a fundamental measure of the interaction radius . for the two - body scattering @xmath117 , an essentially model - independent decomposition is @xmath118 , where @xmath119 come from the size of the @xmath120 transition vertex and @xmath121 comes from the interaction range proper . the values of @xmath119 depend strongly on the excitation energy in the @xmath122 transition . the situation in hadronic diffraction can be summarized as follows ( @xcite , see also the review @xcite ) : in the elastic case , @xmath123 , and for excitation of resonances and of the low - mass continuum states , @xmath124 , which fall into the broad category of _ exclusive _ diffraction , one finds @xmath125 and @xmath126 , here @xmath127 is the hadronic radius squared . in the hadronic elastic scattering @xmath128 4 - 6 gev@xmath66 , and typically @xmath129 gev@xmath66 . but if @xmath130,_i.e . _ , then @xmath131 . for instance , only the target proton size , and @xmath121 , contribute to @xmath116 in the triple - pomeron region of truly _ inclusive _ diffraction @xmath132 summed over all large - mass diffractive states , @xmath133(5 - 10 ) gev@xmath134 . here one finds @xmath1356 gev@xmath66 . in the double high - mass diffraction @xmath136 , when @xmath137 , one is left with @xmath1381.5 - 2 gev@xmath66 ( for the especially illuminating data on double diffraction see @xcite ) . the regge shrinkage of the diffraction , _ i.e. _ , the regge rise of @xmath139 with energy , was seen in all elastic scattering processes ; there is as yet no clear evidence for the shrinkage of the diffraction cone for diffraction dissociation . the remarkable feature of @xmath140 is its universality , because the dependence on the diffracting beam and diffractively produced state drops out ; the same @xmath141 gev@xmath66 is found for @xmath142 and for real photoproduction @xmath143 @xcite . furthermore , the absence of the @xmath37 dependence and the same @xmath144 has been argued to hold for ddis at @xmath145 , _ i.e. _ , at @xmath56 @xcite , which agrees with the first data from zeus lps : @xmath146 gev@xmath66 in ddis for 5 gev@xmath14720 gev@xmath66 @xcite and @xmath148 gev@xmath66 in real photoproduction @xcite . ddis at moderate and large @xmath49 is the tricky one . here the principal issues are : i ) what is the typical dipole size ( scanning radius ) in the @xmath149 transition , and ii ) where is the borderline between _ exclusive _ and _ inclusive _ ddis , _ i.e. _ , what is the relevant excitation scale - the hadronic scale @xmath150 or @xmath151 ? we start with extreme case of exclusive diffraction : elastic production of vector mesons @xmath29 . the major predictions from the color dipole dynamics for elastic production are @xcite : i ) the qcd pomeron is a series of * moving * poles , the regge shrinkage persists at all @xmath37 and for all vector mesons we predict the rise of @xmath116 by @xmath152gev@xmath66 from the cern / fnal to hera energy ( fig . 3 ) . ii ) because of the shrinkage of the photon and the decrease of the scanning radius @xmath31 with @xmath37 , we predict @xmath153 , in agreement with the experiment , see fig . 3 ( a more detailed analysis reveals a nontrivial slowly decreasing contribution to @xmath154 which , however , is numerically small : @xmath155 @xcite ) . iii ) in the proton dissociative vector meson production one expects @xcite @xmath156 plus a contribution from the @xmath157 transition vertex which is small at large @xmath37 , which has been confirmed experimentally . for instance , h1 found @xmath158 for the proton dissociative electroproduction of the @xmath159 at @xmath160 gev@xmath66 @xcite and @xmath161 gev@xmath66 @xcite . iv ) restoration of flavor symmetry is predicted , _ i.e. , _ the values of @xmath162 for different vector mesons must be to a good accuracy equal if production of different vector mesons is compared at equal @xmath163 . evidently , in the elastic production @xmath164(4 - 6 ) gev@xmath66 . the hera results give much too small @xmath165 . a likely explanation is an admixture of proton diffractive @xmath166 , in which @xmath1671.5 - 2 gev@xmath66 . the lps data can clarify the situation . if an anomalously small values of @xmath165 will persist , it will mean we do nt understand the size of protons as probed by gluons and will put in trouble many a models of diffractive production . because of the node in the radial wave function of radially excited @xmath168 , diffractive production @xmath169 is predicted to have many peculiarities : i ) strong suppression of the @xmath170 production ratio @xcite , which has been well confirmed experimentally , ii ) a counterintuitive @xmath171 , although @xmath168 has a larger radius , @xmath172 @xcite . for instance , for the @xmath173 production even the forward dip is possible ! these predictions can be tested soon at hera . ddis at finite @xmath49 is excitation of continuum states , @xmath174 . if excitation of any continuum were an _ inclusive _ diffraction , then one could have expected @xmath175 , as in the large - mass hadronic diffraction and/or large - mass real photoproduction . this is not the case in ddis ; careful analysis of the @xmath40 excitation revealed @xcite nontrivial variation of @xmath176 about @xmath140 , shown in fig . , the excitation scale is set by @xmath151 rather than the quarkonium mass , so that the diffraction slope at fixed @xmath49 must not depend on @xmath37 . second , for @xmath177 one indeed finds an approximately flavor independent @xmath178 , whereas for light flavors @xmath179 is large , similar to the real photoproduction value of @xmath180 , as it has been anticipated in @xcite . third , for heavy flavors the @xmath49-dependence is weaker , because the effective dipole size in the @xmath181 vertex is smaller : @xmath182 . fourth , the decrease of @xmath183 with @xmath49 explains the decrease of @xmath184 at @xmath88 . the predictions @xcite for @xmath185 , shown in fig . 4 , are extremely interesting from the point of view of the exclusive - inclusive duality @xcite . this is a legitimate pqcd region , although a better treatment of the final state @xmath40 interaction is called . in the extreme limit of @xmath88 , _ @xmath186 , the relevant dipole size is given by a small scanning radius @xmath31 and we predict @xmath187 . at @xmath188 we predict a substantial drop of the diffraction slope . the drop is @xmath189 and comes from the rise of the quark helicity changing @xmath190 transition with the momentum transfer . the dropping @xmath116 nicely correlates with the prediction @xcite of a small diffraction slope for @xmath168 production . to test this prediction , it is necessary to experimentally separate the @xmath191 from @xmath192 contributions . otherwise the observed cross section is dominated by @xmath193 for which @xmath194 , which is uninteresting . these predictions can be tested soon with the lps data from zeus and h1 . testing the fundamental pqcd prediction @xcite of the dominance of @xmath193 at large @xmath49 is crucial for the qcd interpretation of ddis . the measurement of @xmath195 with lps is a good substitute for the measurement of @xmath196 by varying the beam energies , which is not possible at the moment . the key is the determination of the quark helicity changing and conserving amplitudes @xmath197 and @xmath198 @xcite . they both contribute to @xmath199 , whereas @xmath200 and @xmath201 . the azimuthal @xmath202 asymmetry is proportional to @xmath203 it provides a stringent test of the pqcd mechanism of ddis . in fig . 5 we show the pqcd born approximation for @xmath204 at @xmath205 and @xmath206 gev@xmath134 ; the higher orders do not change the gross features of @xmath204 @xcite . for the @xmath192-dominated region of @xmath2070.85 - 0.9 , the asymmetry @xmath204 decreases with @xmath37 . in contrast , it rises with @xmath37 for the @xmath191-dominated @xmath2080.9 . the predicted asymmetry is quite large and is measurable with lps . azimuthal asymmetries for @xmath40 dijet production offer more tests of the mechanism of ddis . one interesting prediction @xcite is a dependence of @xmath176 on the angle @xmath209 between the @xmath40 plane and @xmath5 : @xmath210 . for light flavors we predict quite a substantial asymmetry , @xmath211 , with a nontrivial @xmath49 dependence which is different for @xmath191 and @xmath192 . for heavy flavors @xmath212 is smaller . the @xmath49 and flavor dependence of @xmath213 can be tested even without identification of dijets . it is sufficient to study the azimuthal correlation between the recoil protons and any secondary hadron in the @xmath22 debris . the azimuthal correlation between the @xmath8 plane and the @xmath40 dijet plane comes from the @xmath214 term in ( [ eq:1.1 ] ) . the sign of the @xmath215 term is opposite to that for the photon - gluon fusion - dominated jet production in the standard dis , which is an important signature of the qcd mechanism of ddis @xcite . the pqcd mechanism of ddis is well understood . new breakthroughs are expected with the forthcoming lps data . fundamental predictions for the diffraction slope , azimuthal asymmetries and @xmath191 can be tested . abnormally large higher twist effects , a vast variety of pqcd scales , the resulting breaking of diffractive factorization and inapplicability of the dglap evolution to diffractive structure functions have been firmly established theoretically . these findings did not make their way yet into monte carlo codes , which are still based on the discredited factorization approximation . furthermore , final state radiation in ddis and standard dis is different and conclusions from monte carlo comparisons can be quite misleading . the principal remaining task is a derivation of the real and virtual radiative corrections to ddis at moderate and large @xmath49 . we hope to hear more on that at dis98 in brussels . + it is a pleasure to thank the organizers of dis97 for an exciting meeting . nnn is grateful to danny krakauer for helpful comments on the manuscript .	the recent progress in the qcd theory of diffractive dis ( ddis ) is reviewed . we place emphasis on pqcd scales , diffractive factorization breaking , jet and charm production and new fundamental observables which are becoming accessible with the leading proton spectrometers ( lps ) of zeus and h1 . # 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) # 1@xmath0 # 1@xmath0
large scale alignments of quasar polarization vectors were uncovered by hutsemkers ( 1998 ) , looking at a sample of 170 qsos selected from the litterature and confirmed later on a larger sample ( hutsemkers & lamy 2001 ) . the departure to random orientations was found at significance levels small enough to merit deeper investigations . moreover , these alignments seemed to come from high redshift regions , implying that the underlying mechanism might cover physical distances of giga parsecs . a large survey of linear polarization was then started , with the long - term goal to characterize better the polarization properties of quasars , and a short - term goal to investigate the reality of the alignments . this work gives a preliminary analysis of the alignment effect for a total sample of 355 quasars , comprising new polarization measurements from observing runs between 2001 - 2003 and a new comprehensive compilation from the litterature . the sample was chosen from vron - cetty & vron ( 2000 ) and the sloan digital sky survey . among possible targets a preference was given to bright extragalactic sources having a higher probability to show stronger polarization : bal quasars , red quasars . the selection criteria of the sample are detailed elsewhere ( hutsemkers 1998 , hutsemkers et al . 2001 , 2004 ; sluse et al . 2004 ) . in order to avoid possible contaminations from the interstellar medium of the galaxy , only the objects at galactic latitude @xmath1 were selected . above this galactic latitude the interstellar polarization is smaller ( heiles 2000 ) . the observations were carried out at eso la silla 3.6 m efosc2 ( 2001 , 2002 , 2003 ) and paranal vlt / fors1 ( 2003 ) , in broad - band @xmath2 filter , using a wollaston prism and 4 positions of the half - wave plate to derive stokes parameters @xmath3 and @xmath4 ( hutsemkers et al . 2004 ; sluse et al . the photometry was done using the procedure defined by lamy & hutsemkers ( 1999 ) . photon errors on @xmath3 and @xmath4 are @xmath5 . the foreground polarization ( coming from both the instrument and the interstellar medium ) was computed from the observed field stars and subtracted from the quasar polarization . because this was not possible on every frame , we used stokes parameters averaged from the field stars observed during the same run . this average polarization is always small : @xmath6 . final errors were conservatively derived by quadratically adding photon and foreground errors yielding a total error on the qso polarization degree ca . 0.2 - 0.3 % . finally , because the polarization is a positive quantity only , it is debiased according to the wardle & kronberg method ( 1974 ) . to insure a robust measurement of the polarization direction we include only the objects having a polarization degree of @xmath7 and a maximum error in the polarization position angle of @xmath8 . the final sample of polarized quasars comprises 195 quasars in the north galactic pole region ( ngp ) and 160 quasars in the south galactic pole region ( sgp ) . figure 1 shows an aitoff projection of the full sample superimposed on the galactic star polarization map ( heiles 2000 ) . our sample spans the redshift range @xmath9 homogeneously , with a marginally better sampling in the range @xmath10 , mainly due to our bias towards brighter quasars . figure 2 shows a plot of the quasar sample polarization degrees against redshift ( left panel ) . the slight enhancement of highly polarized quasars at lower redshift is attributed to the fact that more radio - loud quasars were observed at low redshifts . overall the quasar polarization degrees do not correlate with redshift which is consistent with previous studies ( berriman et al . 1990 ) . in fig.2 right panel , the histograms of the linear polarization degree of the quasar sample ( black line ) and of the star sample ( heiles 2000 ) selected over the same region ( grey line ) are superimposed . the star polarization histogram shows much smaller polarization degrees ( 94 out of 1996 stars have a polarization degree of 0.5% or higher ) arguing in favor of minor contaminations from the interstellar medium on the quasar sample . in order to detect and assess the significance of the alignments over the complete 3d sample , we used two statistical tests designed for circular data . the detailed procedure is described in hutsemkers ( 1998 ) . the basic idea is to compute for each quasar a statistics @xmath11 taking into account the compactness of a group of @xmath12 neighbors ( in redshift space ) and the dispersion of their polarization direction . the more compact and aligned a group of @xmath12 neighbors , the smaller the statistics @xmath11 . the second statistical test is the andrew - wasserman test designed by bietenholz ( 1986 ) . both statistics can be calculated for the entire sample and the significance of a departure from homogeneity is assessed with monte - carlo reshuffling of the polarization direction over 3-d positions . the significance level is then computed as the probability that a random configuration has a smaller ( @xmath11 test ) or higher statistics ( andrew - wasserman test ) than the observed one . figure 3 right frame shows the logarithm of the significance level of the andrew - wasserman test versus the number of neighbors @xmath12 for three samples of quasars . it clearly shows that , as the size of the sample increases , the alignments are harder and harder to produce from random distributions for any value of @xmath12 . 3 left frame show the polarization map of the high redshift ngp sample ( top ) and the low redshift ngp sample ( bottom ) . the length of the bars is proportional to @xmath13 . objects of both high and low polarization degrees participate to the same alignments . an unambiguous alignment effect of the polarization direction is visible at high redshift , whereas another alignment emerges at low redshift in a different direction . our statistical tests are not invariant under polar coordinates . a way to avoid coordinate dependence is to parallel transport the angles along great circle on the celestial sphere prior to computing the statistics . this was done by jain et al . they confirm the significance of the alignments of the 213-quasar sample over very large regions of the ngp and sgp . the instrumental polarization can be discarded as a source of contamination because , it is very small on efosc2 ( @xmath140.1% ) and because the polarization degree and angle of quasars observed on different instruments are consistent within the errors . the contamination from the galaxy is not a dominant component . indeed , the interstellar polarization is usually much smaller than the polarization of quasars , and the interstellar polarization angles do not follow the quasar alignment directions . moreover it is difficult to explain the redshift dependence of the alignment directions assuming a foreground interstellar screen . if we accept the fact that the alignments are intrinsic to the sources , we have to face correlated polarizations over extreme scales of giga parsecs . exotic sources of polarization on cosmological scales could be invoked , such as pervading exotic particles ( harari & sikivie 1992 ; jain et al . 2004 ) or an intrinsic alignment of the axes of quasar central engines . future theoretical developments need to take these observed correlations into account . the ongoing survey is paramount to characterize the alignments of quasar polarization directions accross the sky and correlate them with the other large - scale dataset ( cmb , galaxy surveys , ... ) . it will also allow us to deepen our knowledge of the linear polarization properties of quasars , and to understand the different source of linear polarization between the different species . we definitely need to increase the sample of polarized quasars from a few hundreds to a few thousands over the next decade to progress significantly in the field . berriman , g. , schmidt , g. d. , west , s. c. , & stockman , h.s . , , 1990 , * 74 * , 869 . bietenholz m. f. , 1986 , , * 91 * , 1249 . harari , d. , & sikivie , p. 1992 b , * 289 * , 67 heiles , c. , 2000 , , * 119 * , 923 . hutsemkers , d. , sluse , d. , cabanac , r. , lamy , h. , quintana , h. , in prep . sluse , d. , hutsemkers , d. , lamy , h. , cabanac , r. , in prep . hutsemkers , d. , 1998 , a&a , * 332 * , 410 . hutsemkers , d. , & lamy , h. , 2001 , a&a , * 367 * , 381 . jain , p. , narain , g. , & sarala , s. , , * 347 * , 394 . lamy , h. & hutsemkers , d. , 1999 , msngr . , * 96 * , 25 . ( erratum : the messenger 97 , 23 ) wardle , j. f. c. & kronberg , p. p. , 1974 , , * 194 * , 249 .	a survey measuring quasar polarization vectors has been started in two regions towards the north and south galactic poles . here , we review the discovery of significant correlations of orientations of polarization vectors over huge angular distances . we report new results including a larger sample of the quasars confirming the existence of coherent orientations at redshifts @xmath0 .
in every major scenario for physics above the @xmath1 scale non - perturbative chiral gauge theory dynamics is expected to play an important role , yet our understanding of this dynamics is very limited . recently we proposed a non - perturbative formulation of chiral gauge theory on the lattice , in the hope that the important features of these theories will be calculable in future computer simulations @xcite . we briefly review it here , focussing on a lattice gauge field interpolation procedure which is the crucial feature of our construction . the fermions are taken to live on a euclidean lattice with spacing @xmath2 . fermion doublers are eliminated by the rome group method of using a gauge non - invariant wilson term @xcite . ( the nielsen - ninomiya theorem tells us that we must break chiral gauge invariance to eliminate the fermions doublers . ) the central problem is then to recover gauge invariance in the continuum limit . before summing over gauge fields this can be arranged in a fairly simple way for any @xmath3-@xmath4 theory @xcite . however , once gauge fields are integrated over , new divergences can lead to uncontrollable violations of gauge invariance . a solution to this problem is to cut off the gauge field momenta by a scale @xmath5 ( @xmath6 in the continuum limit . ) @xcite gives detailed lattice power - counting arguments to show that this can be achieved by obtaining the @xmath2-lattice gauge fields as an interpolation of gauge fields living on a lattice of spacing @xmath7 , which are summed over using the standard wilson action . in practice it is possible that @xmath8 may not have to be too large for computing the properties of low - lying states @xcite . ( for another possible way of cutting off gauge field momenta see ref . @xcite . ) it is sufficient for the interpolation procedure to satisfy the following properties @xcite , most easily stated by imagining interpolating the @xmath9-lattice link variables , @xmath10 , all the way to the continuum to give gauge fields , @xmath11 $ ] . ( i ) transverse continuity : the interpolation describes a differentiable continuum gauge field @xmath12 each @xmath9-lattice hypercube , whose transverse components are continuous across hypercube boundaries ( the longitudinal components can jump ) . ( ii ) lattice spacetime symmetries should be respected . ( iii ) gauge covariance : a @xmath9-lattice gauge transformation changes the interpolation only by a continuous gauge transformation . ( iv ) locality : the gauge - invariant behavior of the interpolation should depend locally on @xmath10 , in particular it is sufficient if the trace of any continuum wilson loop depends only on the @xmath10 lying on @xmath9-hypercubes through which the wilson loop passes . we have detailed such an interpolation procedure for non - abelian gauge fields in four dimensions in ref . below , we describe the more transparent case of interpolating @xmath0 gauge fields in @xmath13 dimensions , following a procedure which readily generalizes to the non - abelian case . for simplicity we deal with the case @xmath14 ( the continuum ) and work in units where @xmath15 . in order to maintain transverse continuity it is helpful to build the interpolation from the lowest dimensional sublattices up . we therefore begin by interpolating the link variables , @xmath16 along the points of each plaquette edge . ( we are neglecting the measure - zero set of lattice fields where at least one of the link variables equals exactly @xmath17 . ) the simplest such interpolation is @xmath18 note that parallel transport along the links agrees between the lattice and the continuum fields . we now @xmath19 to interpolate the lattice field into a plaquette interior in such a way as to agree with the above interpolation on the plaquette edges . in order to satisfy locality we try to do this interpolation for each plaquette , using as input only its bounding link variables . in order to satisfy gauge covariance the strategy is to do a @xmath20 gauge transformation on the bounding links of the plaquette which put the link variables into a complete axial gauge . thus all gauge equivalent lattice fields on the plaquette edges are taken to the same gauge - fixed field ( ` almost ' the same in the non - abelian case ) , which we denote by @xmath21 . this lattice configuration will then be smoothly interpolated to the plaquette interior to give a continuum field @xmath22 . we will then try to find a smooth gauge transformation inside the plaquette which makes the result agree with the one - dimensional edge interpolation , eq . ( 2 ) . at this last stage we will fail , but in a way which we can understand and then correct . in detail , let us fix some plaquette and use @xmath23 coordinates @xmath24 for the vertices of the plaquette . the lattice gauge transformation , @xmath25(z ) = u_1(s)^{z_1 } u_2(s + z_1 \hat{1})^{z_2},\ ] ] takes @xmath26 to @xmath27 this lattice field is easily interpolated into the plaquette interior , @xmath28 where @xmath29 are local continuum coordinates for the plaquette interior . the problem is now to find a @xmath30 gauge transformation @xmath31 which takes @xmath22 to a gauge field agreeing with eq . ( 2 ) , so that we can be assured of transverse continuity across plaquette boundaries . in fact it is not hard to see that this demand essentially fixes @xmath31 on the plaquette boundary to be @xmath32 thereby specifying a map from the plaquette boundary ( topologically a circle ) to @xmath0 ( topologically also a circle ) . we can therefore associate a topological winding number ( integer ) to this map for each plaquette . unless this winding is zero , @xmath31 can not be continuously extended from the plaquette boundary to the interior , and we are stuck . it is simple to show that the winding number associated to the plaquette at @xmath33 , @xmath34 , is equal to @xmath35,\ ] ] ( @xmath36 \equiv$ ] nearest integer to @xmath37 ) and is generically non - zero . the way out of this impasse is to generalize the edge interpolation to @xmath38 where the @xmath39 are integer - valued and do not affect agreement of parallel transport between the link variables and the interpolation . ( in the notation of ref . @xcite , @xmath40 . ) if these integers are chosen to satisfy @xmath41 it is easy to show that this new definition differs from the original by a gauge transformation defined on the plaquette boundary with winding number @xmath42 . therefore the gauge transformation which makes @xmath22 agree with eq . ( 9 ) , @xmath43 , has winding number zero and can be smoothly extended to the plaquette interior , allowing us to get an interpolation @xmath44 . a simple choice for this extension of @xmath43 yields for @xmath45 , @xmath46 if @xmath47 then in fact there is no consistent solution to eq . ( 9 ) the reason is that the boundary conditions for the interpolation then correspond to a continuum configuration with topological charge @xmath47 , which can not be represented by a single smooth periodic gauge field . while configurations with non - zero topological charge are physically important , we do not need them in our proposal for lattice chiral gauge theory , because their effects can be inferred from the sector with zero topological charge using cluster decompostion for the full theory . for @xmath48 eq . ( 9 ) has many solutions and it is simple to pick one @xcite . let us check the four central requirements for a successful interpolation . ( i ) it is straightforward to see that eq . ( 10 ) defines a transversely continuous gauge field . ( ii ) even though we had to pick the axes of our complete gauge fixing somehow , the @xmath49-@xmath50 behavior of our interpolation is covariant under lattice translations and rotations , though the gauge dependent form is not . to see this in the @xmath0 case is easy , since from eqs . ( 10 , 9 ) the continuum field strength is a constant in each plaquette and is just the logarithm of the plaquette field strength ( with absolute value less than @xmath51 ) . for the same reason ( iii ) and ( iv ) are also obvious , the non - locality in choosing the @xmath52 does not infect the gauge invariant part of the interpolation ( ie . the field strength for the @xmath0 case ) . while non - abelian interpolation in four dimensions is technically more complicated , the basic steps are the same . the topological obstruction to making higher dimensional interpolations agree with lower dimensional ones now occurs in four dimensions , because the smallest non - abelian group , @xmath53 , is topologically the 3-sphere as is a hypercube boundary . the resolution of the problem generalizes the 2-d @xmath0 case . one extra complication not seen in two dimensions is that the choice of axes for the complete gauge fixing ( important for maintaining gauge covariance ) @xmath54 lead to a breaking of lattice rotational covariance in the gauge - invariant behavior of the interpolation . to repair this one needs to allow the orientation of the axes in each hypercube to be different and to determine this orientation by the gauge - invariant behavior of the link field @xmath10 itself . then if @xmath10 is rotated , so do the axes . @xcite . the interpolation in ref . @xcite differs in that the authors directly interpolate @xmath31 instead of @xmath43 , thus getting singular gauge fields whenever any @xmath34 is non - zero . such singular gauge fields are unsuitable for our formulation of chiral gauge theories . their interpolation also breaks lattice rotational covariance in the non - abelian case . 9 p. hernndez and r. sundrum , nucl . b455 ( 1995 ) 287 . p. hernndez and r. sundrum , nucl . phys . b472 ( 1996 ) 334 . a. borrelli et . al . , nucl . b333 ( 1990 ) 335 . m. gockeler , g. schierholz , nucl.phys . b(proc.suppl ) 29b , c ( 1992 ) 114 ; nucl.phys . b(proc.suppl ) 30 ( 1993 ) 609 ; g.t . bodwin and e.v . kovacs , nucl.phys . b(proc.suppl ) 30 ( 1993 ) 617 . s.a . frolov and a.a . slavnov , nucl . b411 ( 1994 ) 647 ; a.a . slavnov , phys.lett . b319 ( 1993 ) 231 . m. gckeler , a. kronfeld , g. schierholz and u.j.wiese , nucl . b404 ( 1993 ) 839 .	the importance of lattice gauge field interpolation for our recent non - perturbative formulation of chiral gauge theory is emphasized . we illustrate how the requisite properties are satisfied by our recent four - dimensional non - abelian interpolation scheme , by going through the simpler case of @xmath0 gauge fields in two dimensions .
considerable interest has recently been focused on the optomechanical system as an excellent candidate for studying the transition of a macroscopic degree of freedom from the classical to the quantum regime . this system also provides novel routes for practical applications such as detection and interferometry of gravitational waves@xcite and quantum limited displacement sensing @xcite . the standard and simplest setup of this system is a fabry - perot cavity in which one of the mirrors is much lighter than the other , so that it can move under the effect of the radiation pressure force . state - of - the - art technology allows experimental demonstration of cooling of the vibrational mode of the mechanical oscillator to its ground state@xcite and strong optomechanical coupling between the vibrational mode of the mechanical oscillator and cavity field@xcite . this coupling is intrinsically nonlinear since the length of the cavity depends upon the intensity of the field in analogous way to the optical length of a kerr material@xcite . therefore it enables pondermotive squeezing of the field@xcite , photon blockade @xcite , generation of nonclassical states of the mechanical and optical mode @xcite , optical bistability@xcite and phonon - photon entanglement in the bistable regime @xcite . besides this intrinsic nonlinearity , the presence of an optical parametric amplifier ( opa ) @xcite or the optical kerr medium @xcite inside the cavity has opened up a new domain for combining nonlinear optics and optomechanics towards the enhancement of quantum effects . it has been predicted @xcite that the presence of an opa in a single - mode fabry - perot cavity causes a strong coupling between the oscillating mirror and the cavity mode resulting from increasing the intracavity photon number . also , when the optomechanical cavity contains an optical kerr medium with strong @xmath0 nonlinearity , the photon - photon repulsion and the reduction of the cavity photon fluctuations provide a feasible route towards controlling the dynamics of the micromirror@xcite . on the other hand , the interaction of a single - mode field with both a kerr medium and a parametric nonlinearity is a well - known quantum optical model which has been proposed for the generation of nonclassical states of the cavity field @xcite . here , we consider the interaction of a single - mode field with both a weak kerr medium and a parametric nonlinearity in an intrinsically nonlinear optomechanical system . in particular , we investigate the multistability , intensity , back - action cooling and stationary optomechanical entanglement . it turns out that the mean intracavity photon number , in addition to bistability , exhibits tristability for a certain range of the parameters which can be controlled by the intrinsic and external nonlinearities in the system . then , we investigate the effect of kerr - down conversion nonlinearity on the back - action ground state cooling of the mirror based on the covariance matrix and identify the modified optimal regime for cooling . we will show that the lower bound of the resolved sideband regime and the minimum attainable phonon number can be less than that of a bare cavity by controlling the parametric nonlinearity and the phase of the field driving the opa . also , the weak kerr nonlinearity can be used to extend the domain of the stability to the desired range of the effective detuning in which the effective temperature of the system minimizes . then we show that for a fixed effective detuning , in spite of the bare cavity , one of the stability parameters ( the counterpart of the bistability parameter ) is a nonlinear function of the input power , allows to approach significant entanglement simultaneously with the ground state cooling of the mirror . in the last part of the paper we shall focus on the generation of stationary entanglement in the presence of the nonlinearity and show that not only the entanglement is not a monotonic function of the optomechanical coupling strength , but also the two stability parameters of the system are the key parameters for maximizing the degree of entanglement . based on these results we show that in the tristable regime , for the first and second branches the degree of entanglement is maximum at the end of the branches , while for the third branch the phonon - photon entanglement is null . as is shown in fig . [ fig : fig1 ] , we consider a kerr - down conversion optomechanical system composed of a degenerate opa and a nonlinear kerr medium placed within a fabry - perot cavity formed by a fixed partially transmitting mirror and one movable perfectly reflecting mirror in equilibrium with a thermal bath at temperature @xmath1 . the movable mirror is free to move along the cavity axis and is treated as a quantum mechanical harmonic oscillator with effective mass @xmath2 , frequency @xmath3 and energy decay rate @xmath4 where @xmath5 is the mechanical quality factor . the cavity field is coherently driven by an input laser field with frequency @xmath6 and amplitude @xmath7 through the fixed mirror . furthermore , the system is pumped by a coupling field to produce parametric oscillation and induce the kerr nonlinearity in the cavity . in our investigation , we restrict the model to the case of single - cavity and mechanical modes@xcite . the single cavity - mode assumption is justified in the adiabatic limit , i.e. , @xmath8 in which @xmath9 is the speed of light in vacuum and @xmath10 is the cavity length in the absence of the cavity field . we also assume that the induced resonance frequency shift of the cavity and the kerr medium are much smaller than the longitudinal - mode spacing in the cavity . furthermore , one can restrict to a single mechanical mode when the detection bandwidth is chosen such that it includes only a single , isolated , mechanical resonance and mode - mode coupling is negligible . under these conditions , the total hamiltonian of the system in a frame rotating at the laser frequency @xmath6 can be written as @xmath11 where [ hamil ] @xmath12 the first two terms in @xmath13 are , respectively , the free hamiltonians of the cavity field with annihilation(creation ) operator @xmath14 and the movable mirror with resonance frequency @xmath3 and dimensionless position and momentum operators @xmath15 and @xmath16 . the third term describes the optomechanical coupling between the cavity field and the mechanical oscillator due to the radiation pressure force with coupling constant @xmath17 , and the last term in @xmath13 describes the driving of the intracavity mode with the input laser . also , the two terms in @xmath18 describe , respectively , the coupling of the intracavity field with the opa and the kerr medium ; @xmath19 is the nonlinear gain of the opa which is proportional to the pump power driving amplitude , @xmath20 is the phase of the field driving the opa , and @xmath21 is the anharmonicity parameter proportional to the third order nonlinear susceptibility @xmath22 of the kerr medium . the input laser field populates the intracavity mode through the partially transmitting mirror , then the photons in the cavity will exert a radiation pressure force on the movable mirror . in a realistic treatment of the dynamics of the system , the cavity - field damping due to the photon - leakage through the incomplete mirror and the brownian noise associated with the coupling of the oscillating mirror to its thermal bath should be considered . using the input - output formalism of quantum optics@xcite , we can consider the effects of these sources of noise and dissipation in the quantum langevin equations of motion . for the given hamiltonian ( [ h ] ) , we obtain the following nonlinear equations of motion [ langevin ] @xmath23 where @xmath24 is the cavity decay rate through the input mirror and @xmath25 is the input vacuum noise operator characterized by the following correlation functions @xcite [ noise1 ] @xmath26 the brownian noise operator @xmath27 describes the heating of the mirror by the thermal bath at temperature @xmath28 and is characterized by the following correlation function @xcite @xmath29d\omega,\ ] ] where @xmath30 is the boltzmann constant . we are interested in the steady - state regime and small fluctuations with respect to the steady state . thus we obtain the steady - state mean values of @xmath16 , @xmath15 and @xmath31 as @xmath32 @xmath33 where @xmath34 denotes the new equilibrium position of the movable mirror , @xmath35 , and @xmath36 is the effective detuning of the cavity which includes both the radiation pressure and the kerr medium effects . it is obvious that the optical path and hence the cavity detuning are modified in an intensity - dependent way . the first modification which is a mechanical nonlinearity , arises from the radiation pressure - induced coupling between the movable mirror and the cavity field and the second modification comes from the presence of the nonlinear kerr medium in the optomechanical system . since the mean intracavity photon number in the steady state @xmath37 satisfies a third - order equation , it can have three real solutions and hence the system may exhibit multistability for a certain range of parameters . the multisolution region exists for intracavity intensity values between @xmath38 and @xmath39 with @xmath40 the multistability of the solution fails if @xmath41 and requires @xmath42 . to study the effect of the presence of the kerr - down conversion nonlinearity on the steady - state response of the optomechanical system we consider a cavity with length @xmath43 mm and decay rate @xmath44 which is driven by a laser with @xmath45 nm . the mass , the mechanical frequency and the damping rate of the oscillating end mirror are @xmath46 ng , @xmath47mhz , @xmath48 hz and the environment temperature is @xmath49mk . this set of parameters is close to several optomechanical experiments@xcite . in fig . [ fig : as ] we plot the mean intracavity photon number as a function of the bare detuning @xmath50 for input power @xmath51mw for various values of @xmath21 ( fig . [ fig : as]a ) , @xmath19 ( fig . [ fig : as]b ) , and @xmath20(fig . [ fig : as]c ) . figure [ fig : as]a shows that for @xmath52 ( @xmath53 ) the presence of both intrinsically optomechanical and kerr nonlinearities shift the center of the resonance while the curve is nearly lorentzian . however , for @xmath54 the resonance frequency of the cavity shifts to the lower values and the third order polynomial equation for @xmath55 has three real roots for @xmath56 . the frequency shift of the cavity mode and multi solutions due to the kerr nonlinearity can be reduced or compensated by the term @xmath57 which acts to shift the cavity resonant frequency to the right for @xmath58(figs . [ fig : as]b , c ) . : ( a ) for different values of the anharmonicity parameter@xmath21 with @xmath59 and @xmath60 , ( b ) for different values of the parametric nonlinearity @xmath19 with @xmath61 hz and @xmath60 , and ( c ) for different values of @xmath20 with @xmath62 and @xmath63 hz . , title="fig:",width=2 ] : ( a ) for different values of the anharmonicity parameter@xmath21 with @xmath59 and @xmath60 , ( b ) for different values of the parametric nonlinearity @xmath19 with @xmath61 hz and @xmath60 , and ( c ) for different values of @xmath20 with @xmath62 and @xmath63 hz . , title="fig:",width=2 ] : ( a ) for different values of the anharmonicity parameter@xmath21 with @xmath59 and @xmath60 , ( b ) for different values of the parametric nonlinearity @xmath19 with @xmath61 hz and @xmath60 , and ( c ) for different values of @xmath20 with @xmath62 and @xmath63 hz . , title="fig:",width=2 ] in order to investigate the dynamics of the system , we need to study the dynamics of small fluctuations near the steady state . we assume that the nonlinearity in the system is weak and decompose each operator in eq.([langevin ] ) as the sum of its steady - state value and a small fluctuation with zero mean value , @xmath64 inserting the above linearized forms of the system operators into eq.([langevin ] ) , and defining the cavity field quadratures @xmath65 and @xmath66 and the input noise quadratures @xmath67 and @xmath68 the linearized quantum langevin equations for the fluctuation operators can be written in the compact matrix form @xmath69 where @xmath70 is the vector of fluctuations , @xmath71 is the vector of noise sources and the matrix @xmath72 is given by @xmath73 where @xmath74 is the enhanced optomechanical coupling rate , @xmath75 ( @xmath76 ) is the real ( imaginary ) part of @xmath77 , and @xmath78 . here , we concentrate on the stationary properties of the system . for this purpose , we should consider the steady - state condition governed by eq.([matrixform ] ) . the system is stable only if the real part of all eigenvalues of the matrix @xmath72 are negative , which is also the requirement of the validity of the linearized method . the parameter region in which the system is stable can be obtained from the routh - hurwitz criterion@xcite , which gives the following three independent conditions : [ routh-1 - 3 ] @xmath79 the violation of the third condition , @xmath80 , indicates instability in the region @xmath81 . for the bare cavity ( @xmath82 ) this condition reduces to the instability in the domain of blue - detuned laser . within this frequency range , the effective mechanical damping rate becomes negative and self - sustained oscillations set in@xcite . the violation of the second condition , @xmath83 , indicates instability in the region @xmath84 . for a bare cavity this condition cause a bistability of the system@xcite . there is an additional stability condition given by @xmath85 , which is always satisfied for the bare cavity , and gives the condition for the threshold for parametric oscillation . accordingly , for @xmath84 we can define the following stability parameters @xmath86 for @xmath82 , @xmath87 reduces to the well known `` bistability parameter '' @xcite . one of the main features arising from the kerr - down conversion nonlinearity is the appearance of three stable states for the mirror . figure [ fig : tristability ] shows the hystersis loop for the intracavity mean photon number when @xmath56 ( blue - detuned laser ) . in this figure the unstable solutions are represented by dashed lines . it shows that depending on the value of @xmath20 the steady - state response of the mirror can be monostable , bistable and tristable . from an experimental point of view , controllable triple - state switching is possible practically by adding a pulse sequence onto the input field@xcite . such tristability can be used for all - optical switching purposes function as memory devices for optical computing and quantum information processing . versus the input power @xmath88 at @xmath89.the solid and dashed lines correspond to the stable and unstable branches , respectively . the parameters are : @xmath90 , @xmath91 hz , @xmath92 . other parameters are the same as those in fig . [ fig : as ] , width=2 ] due to the linearization method and the gaussian nature of the noise operators the asymptotic steady state of the quantum fluctuations is a zero mean gaussian state . as a consequence , the steady state can be fully characterized by the covariance matrix ( cm)@xmath93 . this formalism provides a unified framework for exploring both cooling of the mechanical oscillator and phonon - photon entanglement . when the stability conditions of eq.([routh-1 - 3 ] ) are fulfilled , we can solve eq.([matrixform ] ) for the @xmath94 stationary correlation matrix @xmath93@xcite @xmath95 where the elements of the correlation matrix @xmath93 are defined as @xmath96 and @xmath97diag@xmath98 $ ] is the diagonal diffusion matrix in which we used the following approximation @xmath99 where @xmath100^{-1}$ ] . the lyapunov eq . ( [ lyapunov ] ) for the steady state cm can be straightforwardly solved . however , the general exact expression is too cumbersome and will not be reported here . it has been shown @xcite that within each optomechanical cavity the radiation pressure coupling can lead the system into a stationary state with genuine quantum features , including cooling of the resonator towards the ground state and photon - phonon entanglement . it is therefore interesting to investigate these quantum features in the presence of kerr - down conversion nonlinearity . the cm , @xmath93 , contains all the information about the steady state . in particular , the mean energy of the mechanical oscillator is given by @xmath101 where @xmath102 is the mean effective excitation number of the mechanical mode corresponding to an effective mode temperature @xmath103 . the ground - state cooling is approached if @xmath104 , or @xmath105 . moreover , the photon - phonon entanglement can be quantified by the logarithmic negativity @xmath106 @xcite @xmath107,\ ] ] where @xmath108^{1/2}$ ] is the lowest symplectic eigenvalue of the partial transpose of the cm with @xmath109 , and we have used the @xmath110 block form of the cm @xmath111 with @xmath112 associated to the oscillating mirror , @xmath113 to the cavity mode , and @xmath114 describing the optomechanical correlations . we now focus on numerical examples to see how the kerr - down conversion nonlinearity affects the ground - state cooling of the movable mirror . figure [ fig : t1 ] shows the variation of the effective temperature @xmath115 with @xmath116 and @xmath19 for the red - detuned laser(@xmath117 ) and in the monostable regime . in this figure we have assumed the good - cavity limit , @xmath118 , and the reservoir temperature to be @xmath119 . the effective mode temperature of the bare cavity ( red dashed line ) is plotted for comparison . it shows that the parametric process inside the cavity leads to lower temperature . to illustrate this explicitly , we can solve eq . ( [ lyapunov ] ) in the limit of large mechanical quality factor and low - temperature environment ( i.e. , @xmath120 and @xmath121 ) and obtain the position ( @xmath122 ) and momentum ( @xmath123 ) variances and @xmath102 by using @xmath124 . the results are as follows @xmath125 and for @xmath126 , @xmath127 where @xmath128 . for a bare cavity @xmath102 reduces to eq . ( 6 ) in @xcite . the minimum attainable phonon number can be achieved for @xmath129 , @xmath130 in agreement with eq . ( 7 ) in @xcite for @xmath131 . as can be seen , when @xmath132 the lower bound of the resolved sideband regime(@xmath133 ) and @xmath134 can be less than that of a bare cavity . for the data used in fig . [ fig : t1 ] , and for @xmath59 , @xmath135 we have @xmath136 and the minimum effective temperature is about @xmath137mk at @xmath138 . also , for @xmath139 and @xmath140 we have @xmath141 and the minimum effective temperature is about @xmath142mk at @xmath143 . according to eq . ( [ nmin ] ) it seems that @xmath144 does not depend on the anharmonicity parameter @xmath21 , but it should be noted that even if the effect of kerr nonlinearity on @xmath87 can be eliminated , the above analysis is carried out for the limit of low - temperature environment ( @xmath121 ) . in a realistic case the thermal noise will also be present , which in turn modifies the minimum attainable temperature . figures [ fig : t - kerr](a ) and ( b ) show the variation of effective temperature with @xmath145 for different values of @xmath21 and for two different environment temperatures @xmath146mk and @xmath147mk . for @xmath139 and @xmath148 the minimum temperature is achieved for the optimal detuning @xmath149 . as can be seen , the system is unstable in this range for @xmath150 and @xmath61 . for @xmath151 and @xmath152 the stability domain is extended to the desired range and the effective temperature is increased with increasing @xmath21 . as shown in fig . [ fig : t - kerr](b ) this heating effect is smaller for lower environment temperature @xmath153mk . versus @xmath154 for different values of @xmath19 . the red dashed curve corresponds to the bare cavity . the parameters are : @xmath91hz , @xmath135 for @xmath59 and @xmath140 for @xmath139 , @xmath146mk , @xmath155 mw , and @xmath118 . other parameters are the same as those in fig . [ fig : tristability ] . , width=2 ] versus @xmath145 for environment temperature ( a ) @xmath146(mk ) and ( b ) @xmath153(mk ) for different values of @xmath21.the red dashed curve corresponds to the cavity containing only the gain nonlinearity . the parameters are:@xmath139 , @xmath148 , and @xmath155 mw . other parameters are the same as those in fig . [ fig : tristability ] and the stability parameter is fixed to be @xmath156 . , title="fig:",width=2 ] versus @xmath145 for environment temperature ( a ) @xmath146(mk ) and ( b ) @xmath153(mk ) for different values of @xmath21.the red dashed curve corresponds to the cavity containing only the gain nonlinearity . the parameters are:@xmath139 , @xmath148 , and @xmath155 mw . other parameters are the same as those in fig . [ fig : tristability ] and the stability parameter is fixed to be @xmath156 . , title="fig:",width=2 ] versus @xmath154 and input power @xmath88 for @xmath151 . other parameters are the same as those in fig . [ fig : t - kerr](a ) . , width=2 ] and ( b ) the logarithmic negativity versus the input power @xmath88 for @xmath157 , @xmath158 , @xmath159 and @xmath91 . other parameters are the same as those in fig . [ fig : t1 ] . , title="fig:",width=2 ] and ( b ) the logarithmic negativity versus the input power @xmath88 for @xmath157 , @xmath158 , @xmath159 and @xmath91 . other parameters are the same as those in fig . [ fig : t1 ] . , title="fig:",width=2 ] also , in fig . [ fig : t - kerr ] the dependence of effective temperature on the kerr nonlinearity has been illustrated for fixed values of stability parameter @xmath87 . it should be emphasized , however , that the controllable ground state cooling of the vibrational mode is possible only if the limit @xmath126 is reachable in the presence of the nonlinear medium . figure [ fig : eta1 ] shows the variation of the parameter @xmath87 versus the input power @xmath88 and the normalized effective detuning @xmath154 for the data of fig . [ fig : t - kerr](a ) and for @xmath151 . as can be seen , @xmath160 and the required condition holds actually very well . also , fig . [ fig : eta1 ] shows that for a fixed effective detuning , in contrast to the case of bare cavity @xcite , @xmath87 is not a linear function of the input power @xmath88 . this feature arises from the kerr nonlinearity ( the term @xmath161 in @xmath162 and @xmath163 in eq.([eta1,2 ] ) ) and allows to approach significant entanglement simultaneously with the ground state cooling of the mirror ( since one can enhance optomechanical coupling by increasing the input power while @xmath87 remains approximately unaffected ) . as an example , [ fig : ent1 ] represents the optomechanical entanglement and the effective temperature for @xmath158 , @xmath91 , and @xmath157 as functions of the input power . it shows that the minimum value of the effective temperature @xmath115 and the maximum value of entanglement is achieved in the same range of the input power . nonetheless , we will show that as in the case of bare cavity , entanglement and cooling are different phenomena and generally are optimized in different regimes . we study the effects of @xmath20 , @xmath19 and @xmath21 on @xmath164 separately to find the regime of maximal phonon - photon entanglement . we first study the behavior of @xmath164 when the phase of the field driving the opa , i.e. , @xmath20 , is varied . figure [ fig : theta ] shows that entanglement increases with decreasing the phase of the field driving the opa . this entanglement increment is related to the increasing of the photon number and the optomechanical coupling rate @xmath165 ( see fig . [ fig : as](c ) ) . for different values of @xmath20 . the parameters are : @xmath166mw , @xmath158 , @xmath91 . other parameters are the same as those in fig . [ fig : t1 ] . , width=2 ] and input power @xmath88 for ( a ) @xmath59 , and ( b ) @xmath90 . the parameters are : @xmath159 , @xmath91 . other parameters are the same as those in fig . [ fig : t1 ] . , title="fig:",width=2 ] and input power @xmath88 for ( a ) @xmath59 , and ( b ) @xmath90 . the parameters are : @xmath159 , @xmath91 . other parameters are the same as those in fig . [ fig : t1 ] . , title="fig:",width=2 ] figure [ fig : ent - g ] shows the entanglement as a function of the normalized effective detuning @xmath154 and laser power @xmath88 for ( a)@xmath59 and ( b ) @xmath90 . it shows that the maximum possible values of the logarithmic negativity occur at low values of the input power and increase with increasing the gain nonlinearity . it is notable that here the entanglement increment is not just related to the increasing of the optomechanical coupling rate @xmath165 . in table [ table:1 ] we summarize the maximum values of the logarithmic negativity @xmath164 and other parameters associated with it for input powers of @xmath167 mw and @xmath155mw related to fig . [ fig : ent - g ] . in this table one can see that although the optomechanical coupling strength is larger for @xmath59 for both values of the input power but @xmath164 is smaller . this feature suggests to examine other key parameters that determine the entanglement behavior . according to the data given in table [ table:1 ] for both @xmath59 and @xmath90 the maximum values of the entanglement are obtained for @xmath168 very close to unity but the effective detuning and stability parameter @xmath87 are smaller for @xmath90 . this result is in agreement with that obtained in refs . @xcite for a bare cavity , showing that the entanglement becomes maximal for @xmath169 . it is intresting to note that in contrast to bare cavity for @xmath167mw and @xmath59 the maximum value of the logarithmic negativity is obtained for the minimum value of @xmath87 , since at the end of the stability region @xmath170 and @xmath171 . .calculated logarithmic negativities , normalized effective detunings , normalized optomechanical couplings , and the two stability parameters for the input powers @xmath167mw and @xmath155mw in fig . [ fig : ent - g ] . [ cols="^,^,^,^,^,^,^,^",options="header " , ] figure [ fig : ent - kerr ] illustrates the behavior of @xmath164 when the anharmonicity parameter , @xmath21 , is varied . it shows that for the cavity with only the gain nonlinearity the amount of achievable optomechanical entanglement at the steady state is seriously limited by the stability condition . the kerr nonlinearity makes the stability region larger , allows overcoming this limitation . as before , the reduction of the entanglement for @xmath91 ( in comparison to @xmath151 ) is related to the increment of the stability parameter @xmath87 . for @xmath150 , @xmath151 , and @xmath91 . the parameters are : @xmath159 , @xmath90 , and @xmath172mw . other parameters are the same as those in fig . [ fig : t1 ] . , width=3 ] until now we have studied the entanglement in the good cavity limit ( @xmath173 ) and have chosen the values of the input power and the detuning far from the multistability region . it is also interesting to examine the entanglement in the bad cavity limit and in the tristable regime . figure [ fig : ent - tri]shows the entanglement and the stability parameter @xmath87 as functions of the laser input power @xmath88 in the tristable regime for the data of fig . [ fig : tristability ] . it shows that for the first and second branches the entanglement is maximum at the end of the branches , while for the third branch the phonon - photon entanglement is null . this result might be interpreted as arising from the kerr - induced shift of the resonance frequency of the cavity ( for the third branch this shift is more than @xmath174 ) . also , it can be seen that at the end of each stable branch @xmath175 and the entanglement is found only in the region with @xmath176 . as functions of the input power @xmath88 corresponding to the three stable branches for @xmath177 in fig . [ fig : tristability ] . the dashed line corresponds to the unstable region.,title="fig:",width=3 ] as functions of the input power @xmath88 corresponding to the three stable branches for @xmath177 in fig . [ fig : tristability ] . the dashed line corresponds to the unstable region.,title="fig:",width=3 ] it should be noted that in order to stay within the range of validity of the linearization approximation , we have been assured that the condition @xmath178 is satisfied in all the results given above . in conclusion , we have studied the interaction of a single - mode field with both a weak kerr medium and a parametric nonlinearity in an intrinsically nonlinear optomechanical system . we have investigated the stability behavior of the intracavity mean photon number , the intensity , cooling and stationary entanglement . we have found that the combination of the nonlinearities leads to a shift of the resonance frequency of the cavity toward the lower values and appearance of three real roots for the steady - state response of the system . we have derived the general condition for tristability in the system . furthermore , we have studied the cooling and stationary entanglement by using the covariance matrix formalism . in particular , we have shown that the lower bound of the resolved sideband regime and the minimum attainable phonon number can be less than that of a bare cavity by controlling the parametric nonlinearity and the phase of the field driving the opa . the weak kerr nonlinearity can be used to extend the domain of the stability to the desired range of the effective detuning in which the effective temperature of the system is minimized . also , the kerr nonlinearity modifies the behavior of the stability parameter and allows to approach significant entanglement simultaneously with the ground state cooling of the mirror . in the investigation of the degree of stationary entanglement between the cavity and the mechanical modes we have identified four key parameters that affect the behavior of the entanglement . they are effective detuning of the cavity , the optomechanical coupling and the two stability parameters of the system . also , as shown in this paper the present scheme allows to have significant entanglement in the tristable regime for the lower and middle branches which makes the current scheme distinct from the bare optomechanical system . is grateful to m. xiao and r.ghobadi for useful discussions . the authors wish to thank the office of graduate studies of the university of isfahan for their support . v. braginsky and s. p. vyatchanin , phys . a * 293 * , 228 ( 2002 ) . d. rugar , r. budakian , h. j. mamin , and b. w. chui . , nature ( london ) * 430 * , 329 ( 2004 ) . a. d. o connell , m. hofheinz , m. ansmann , r. c. bialczak , m. lenander , e. lucero , m. neeley , d. sank , h. wang , m. weides , j. wenner , j. m. martinis , and a. n. cleland , nature ( london)464 , 697 ( 2010 ) . j. d. teufel , t. donner , d. li , j. w. harlow , m. s. allman , k. cicak , a. j. sirois , j. d. whittaker , k. w. lehnert , and r. w. simmonds , nature ( london ) * 475 * , 359 ( 2011 ) . j. chan , t. p. m. alegre , a. h. safavi - naeini , j. t. hill , a. krause , s. groblacher , m. aspelmeyer , and o. painter , nature ( london ) * 478 * , 89 ( 2011 ) . o. arcizet , p. f. cohadon , t. briant , m. pinard , and a. heidmann , nature ( london ) * 444 * , 71 ( 2006 ) . s. gigan , h. r. bohm , m. paternostro , f. blaser , g. langer , j. b. hertzberg , k. c. schwab , d. bauerle , m. aspelmeyer , and a. zeilinger , nature ( london ) * 444 * , 67 ( 2006 ) . c. h. metzger and k. karrai , nature ( london ) * 432 * , 1002 ( 2004 ) . s. groblacher , k. hammerer , m.r . vanner , and m. aspelmeyer , nature ( london ) * 460 * , 724 ( 2009 ) . z. r. gong , h. ian , yu - xi liu , c. p. sun , and f. nori , phys . a * 80 * , 665801 ( 2009 ) . c. fabre , m. pinard , s. bourzeix , a. heidmann , e. giacobino , and s. reynaud , phys . a * 49 * , 1337 ( 1994 ) ; s. mancini and p. tombesi , phys . rev . a*49 * , 4055 ( 1994 ) . p. rabl , phys . lett . * 107 * , 063601 ( 2011 ) . a. nunnenkamp , k. borkje , and s. m. girvin , phys . * 107 * , 063602 ( 2011 ) . a. dorsel , j. d. mccullen , p. meystre , e. vignes , and h. walther , phys . rev.lett . * 51 * , 1550 ( 1983 ) ; a. gozzini , f. maccarone , f. mango , i. longo , and s. barbarino , j. opt . b * 2 * , 1841 ( 1985 ) . r. ghobadi , a. r. bahrampour , and c. simon , phys . a*84 * , 033846(2011 ) . s. huang and g. s. agarwal , phys . a * 79 * , 013821 ( 2009 ) . a. xuereb , m. barbier , and m. paternostro , phys . rev . a * 86 * , 013809 ( 2012 ) . t. kumar , a. bhattacherjee , and manmohan , phys . a * 81 * , 013835 ( 2010 ) . b. wielinga and g. j. milburn , phys . rev . a * 48 * , 2494 ( 1993 ) . w. leonski , phys . a * 54 * , 3396 ( 1996 ) . c. k. law , phys a * 21 * , 2537 ( 1995 ) . s. mancini , v. i. manko , and p. tombesi , phys . a * 55 * , 3042 ( 1997 ) . s. bose , k. jacobs , and p. l. knight , phys . a * 56,4175 ( 1997 ) . c. w. gardiner and p. zoller , _ quantum noise _ ( springer - verlag , berlin , 1991 ) . v. giovannetti and d. vitali , phys . a 63 * , 023812 ( 2001 ) . l. landau and e. lifshitz , statistical physics ( pergamon , new york , 1958 ) . d. kleckner and d. bouwmeester , nature(london ) * 444 * , 75 ( 2006 ) . d. kleckner et al . , phys . lett . * 96 * , 173901 ( 2006 ) . t. carmon , h. rokhsari , l. yang , t. j. kippenberg , and k. j. vahala , phys . * 94 * , 223902 ( 2005 ) . i. s. gradshteyn and i. m. ryzhik , _ table of integrals , series and products _ ( academic press , orlando , 1980 ) ; a. hurwitz , _ selected papers on mathematical trends in control theory _ , edited by r. bellman and r. kabala ( dover , new york , 1964 ) . f. marquardt , j. g. harris , and s. m. girvin , phys . lett . * 96 * , 103901(2006 ) . m. ludwig , c. neuenhahn , c. metzger , a. ortlieb , i. favero , k . karrai , and f. marquardt , phys . lett . * 101 * , 133903(2008 ) . c. genes , d. vitali , p. tombesi , s. gigan , and m. aspelmeyer , phys.rev.a * 77 * , 033804 ( 2008 ) . j. sheng , u. khadka , and m. xiao , phys . lett . * 109 * , 223906 ( 2012 ) . d. vitali , s. gigan , a. ferreira , h. r. bohm , p. tombesi , a. guerreiro , v. vedral , a. zeilinger , and m. aspelmeyer , phys . lett . * 98 * , 030405 ( 2007 ) . c. genes , a. mari , p. tombesi , and d. vitali , phys . a * 78 * , 032316 ( 2008 ) . a. mari and j. eisert , phys . * 103 * , 213603 ( 2009 ) . f. marquardt , j. p. chen , a. a. clerck , and s. m. girvin , phys . lett.99 , 093902 ( 2007 ) . i. wilson - rae , n. nooshi , w. zwerger , and t. j. kippenberg , phys . rev . lett.99 , 093901 ( 2007 ) . m. paternostro , d. vitali , s. gigan , m. s. kim , c. brukner , j. eisert , and m. aspelmeyer , phys . 99 , 250401 ( 2007 ) . j. m. dobrindt , i. wilson - rae , and t. j. kippenberg , phys . rev . lett.101 , 263602 ( 2008 ) . c. genes , a. mari , d. vitali , and p. tombesi , adv . . phys . * 57 * , 33 ( 2009 ) . g. adesso , a. serafini , and f. illuminati , phys.rev.a*70 * , 022318 ( 2004 ) .	the interaction of a single - mode field with both a weak kerr medium and a parametric nonlinearity in an intrinsically nonlinear optomechanical system is studied . the nonlinearities due to the optomechanical coupling and kerr - down conversion lead to the bistability and tristability in the mean intracavity photon number . also , our work demonstrates that the lower bound of the resolved sideband regime and the minimum attainable phonon number can be less than that of a bare cavity by controlling the parametric nonlinearity and the phase of the driving field . moreover , we find that in the system under consideration the degree of entanglement between the mechanical and optical modes is dependent on the two stability parameters of the system . for both cooling and entanglement , while parametric nonlinearity increases the optomechanical coupling , the weak kerr nonlinearity is very useful for extending the domain of the stability region to the desired range in which the minimum effective temperature and maximal entanglement are attainable . also , as shown in this paper , the present scheme allows to have significant entanglement in the tristable regime for the lower and middle branches which makes the current scheme distinct from the bare optomechanical system .
interference is a major performance - limiting factor in modern wireless communication systems . many interference mitigation strategies have been proposed to improve network spectral efficiency . by allowing partial or full cooperation among interfering base stations , interference can effectively be managed and spectral efficiency can be improved . joint beamforming @xcite and network mimo ( or multicell processing ) @xcite among base stations have been shown to be effective interference mitigation techniques . however , if cooperation among transmitters is not allowed , orthogonal multiple access has been a traditional solution to interference . in a @xmath0-user single - input single - output ( siso ) interference channel ( ic ) , for example , each user can achieve @xmath2 degrees of freedom ( dof ) by time division multiple access . in recent years , interference alignment ( ia ) techniques have received much attention @xcite . the basic concept of ia is to align the interfering signals in a small dimensional subspace . in a @xmath0-user siso ic , @xmath3 dof have been shown to be achievable using ia @xcite . although ia provides a substantial asymptotic capacity gain in interference channels , there are many practical challenges for implementation . ia requires global channel state information at the transmitter ( csit ) , and imperfect channel knowledge severely degrades the gain of ia . in some channel configurations , symbols should be extended in the time / frequency domain to align interfering signals . the high computational complexity is also considered as a major challenge . to ameliorate these difficulties , many ia algorithms have been proposed such as iterative ia @xcite and a subspace ia @xcite . for interference suppression , multiuser diversity can also be exploited by opportunistic user selection for minimizing interference . the interference reduction by multiuser diversity can be enjoyed without heavy burden on global channel knowledge because user selection in general requires only a small amount of feedback @xcite . in this context , opportunistic interference alignment ( oia ) has been recently proposed in @xcite and has attracted much attention . in a 3-transmitter @xmath4 mimo interfering broadcast channel ( ibc ) , the authors of @xcite proved that @xmath5 ( where @xmath6 $ ] ) dof per transmitter is achievable when the number of users scales as @xmath7 . in @xcite and @xcite , a @xmath0-transmitter @xmath8 simo ibc and a @xmath0-transmitter @xmath9 simo ibc have been studied , respectively . for simo interfering multiple access channel ( imac ) constituted by @xmath0-cell uplink channels with @xmath10 transmit antennas and single antenna users , the authors of @xcite showed that @xmath11 dof are achievable when the number of users scales as @xmath12 . in these schemes , user dimensions are used to align the interfering signals ; each transmitter opportunistically selects a user whose interfering signals are most aligned among the users associated with the transmitter . contrary to the conventional opportunistic user selection techniques @xcite , the oia scheme exploits the multiuser dimensions for interference alignment . in this paper , we investigate the optimal role of multiuser diversity for the target dof in the ibc with @xmath0-transmitters and generalize the results of @xcite@xcite . for the @xmath0-transmitter simo ibc , each transmitter selects and serves a single user in its user group consisting of @xmath13 users . once after @xmath0-transmitters select their serving users , a @xmath0-user ic is constructed . each user has @xmath14 antennas less than or equal to the number of interferers , i.e. , @xmath15 . thus , without help of multiuser diversity , interference at each user can not be perfectly removed so that the achievable rate of each transmitter goes to zero as signal - to - noise ratio ( snr ) increases . consequently , the achievable dof per transmitter becomes zero . however , non - zero dof per transmitter is achievable by exploiting multiuser diversity as the number of users increases . since opportunistic user selection can focus on either enhancing the desired signal or decreasing interference , non - zero dof can be obtained by properly enhancing the desired signal strength and reducing interference via user selection . that is , the non - zero dof @xmath16 comprises a dof gain term @xmath17 from the desired signal and a dof loss term @xmath18 caused by interference such that @xmath19 , and the target dof @xmath16 can be obtained by a proper combination of @xmath20 and @xmath21 . however , many questions remain unsolved ; what is the feasible and optimal combination of @xmath22 for the target dof @xmath23 and what is the sufficient number of users for the target dof achieving strategy . we answer these fundamental questions and analytically investigate how the multiuser dimensions can be optimally exploited for the target dof in the ibc . specifically , from geometric interpretation of interfering channels , we define an interference alignment measure that indicates how well interference signals are aligned at each user . using the interference alignment measure , we first consider the @xmath0- transmitters simo ibc and show that the dof gain term @xmath20 can be achieved if the number of users scales in terms of transmit power @xmath24 as @xmath25 and the dof loss term can be reduced to @xmath21 if the number of users scales as @xmath26 . from these results , we find the optimal strategy of exploiting multiuser diversity for the target dof @xmath16 in terms of the required number of users ; the optimal target dof achieving strategies @xmath27 are @xmath28 and @xmath29 for the target dof @xmath30 $ ] and @xmath31 , respectively . we also investigate how the optimal target dof achieving strategy @xmath27 can be realized by practical user selection schemes . then , we extend our results to the @xmath0-transmitter mimo ibc where each transmitter has @xmath32 multiple antennas and serves multiple users with @xmath14 receive antenna each . our generalized key findings are summarized as follows : * for the target dof @xmath33 $ ] , @xmath34 is the optimal target dof achieving strategy that minimizes the required number of users . that is , the multiuser dimensions should be exploited to make the dof loss @xmath35 . the sufficient number of users for this strategy scales like @xmath36 . * for the target dof @xmath37 , @xmath38 is the optimal target dof achieving strategy which minimizes the required number of users . that is , the multiuser dimensions should be exploited to make the dof loss term zero as well as to make the dof gain term @xmath16 . the sufficient number of users for this strategy scales like @xmath39 . the rest of this paper is organized as follows . in section ii , we describe the system model . in section iii , a geometric interpretation of interfering channels is provided , and the interference alignment measure is defined . section iv derives the optimal strategies of achieving the target dof in terms of the required number of users . in section v , we show how various practical user selection schemes exploit multiuser diversity for the target dof and discuss their optimality to achieve the target dof . the system model is extended for the mimo ibc in section vi . numerical results are shown in section vii , and we conclude our paper in section viii . _ notations _ throughout the paper , we use boldface to denote vectors and matrices . the notations @xmath40 , @xmath41 , and @xmath42 denote the conjugate transpose , the @xmath43th largest eigenvalue , and the eigenvector of matrix @xmath44 corresponding to the @xmath43th largest eigenvalue . for convenience , the smallest eigenvalue , the largest eigenvalue , and the eigenvectors corresponding eigenvectors of @xmath44 are denoted as @xmath45 , @xmath46 , @xmath47 , and @xmath48 , respectively . also , @xmath49 , @xmath50 , and @xmath51 indicate the @xmath52 identity matrix , the @xmath53-dimensional complex space , and the set of @xmath54 complex matrices , respectively . our system model is depicted in fig . [ fig : system_model ] . the system corresponds to the interfering broadcast channel ( ibc ) of which capacity is unknown . there are @xmath0 transmitters with @xmath32 transmit antennas each , and each transmitter has its own user group consisting of @xmath13 users with @xmath14 antennas each . first , each transmitter is assumed to have a single antenna , i.e. , @xmath55 , and serves a single user selected in its user group so that @xmath0-transmitter simo ic is opportunistically constituted . the system model with multiple transmit antennas ( i.e. , @xmath56 ) becomes statistically identical with the single transmit antenna model if each transmitter uses a random precoding vector . in section [ sec : extension ] , we extend our system model to the @xmath0-transmitter mimo ibc where each transmitter with multiple antennas serves the multiple users through orthonormal random beams . in this paper , we focus on the cases that the number of receive antennas is smaller than the number of transmitters , i.e. , @xmath57 . otherwise ( i.e. , if @xmath58 ) , each user can suppress all interfering signals through zero - forcing like schemes so that dof one is trivially guaranteed at each transmitter . we also assume that collaboration or information sharing among the transmitters is not allowed . since the user selection at each transmitter is independent of the other transmitters , we only consider the achievable rate of the first transmitter without loss of generality . note that the average achievable rate per transmitter will be same if the configurations of the transmitters are identical . at the first transmitter , the received signal at the @xmath53th user denoted by @xmath59 is given by @xmath60 where @xmath61 is the vector channel from the @xmath62th transmitter to the @xmath53th user whose elements are independent and identically distributed ( i.i.d . ) circularly symmetric complex gaussian random variables with zero means and unit variance . also , @xmath63 is the transmitted signal using random gaussian codebook from the @xmath62th transmitter such that @xmath64 , where @xmath24 is the power budget at each transmitter . also , @xmath65 is a circularly symmetric complex gaussian noise with zero mean and an identity covariance matrix , i.e. , @xmath66 . assuming perfect channel estimation at each receiver , the channel state information @xmath67 is available at the @xmath53th user . the received signal is postprocessed at each user using multiple receive antennas . let @xmath68 be the postprocessing vector of the @xmath53th user such that @xmath69 . then , the received signal after postprocessing becomes @xmath70 to aid user selection at the transmitter , each user feeds one scalar value back to the transmitter . various user selection criteria and corresponding feedback information will be discussed in the following sections . since no information is shared among the transmitters , each transmitter independently selects a single user based on the collected information . let @xmath71 be the index of the selected user at the first transmitter . then , the average achievable rate of the first transmitter is given by @xmath72 we decompose @xmath73 into two terms @xmath74 and @xmath75 such that @xmath76 , which are given , respectively , by @xmath77 then , the achievable dof of the first transmitter becomes @xmath78 we call @xmath79 and @xmath80 as _ dof gain term _ and _ dof loss term _ , respectively . the achievable rate of each transmitter depends on the number of users because multiuser dimensions are exploited for a rate increase . when there are fixed number of users , the achievable rate of each transmitter will be saturated in the high snr region due to interferences because the number of receive antennas at each user is smaller than the number of total transmitters . consequently , the first transmitter can not obtain any dof , i.e. , @xmath81 in this case , both the dof gain term and the dof loss term become one , i.e. , @xmath82 on the other hand , when the transmit power is fixed , the achievable rate of the selected user can increase to infinity as the number of users increases , i.e. , @xmath83 then , how much dof can be achieved when both the number of users and the transmit power increase ? obviously , non - zero dof can be obtained by exploiting multiuser dimensions , and the achievable dof @xmath84 \label{eqn : opportunistic_dof}\end{aligned}\]]will depend on the increasing speeds of @xmath13 and @xmath24 . in this case , dof @xmath85 at the first transmitter comprises the dof gain term @xmath86 and the dof loss term @xmath87 such that @xmath88 , i.e. , @xmath89 we call @xmath90 as a _ target dof achieving strategy _ if @xmath91 for the target dof @xmath16 . since each strategy requires different user scaling , we need to find the optimal dof achieving strategy that exploits multiuser diversity most efficiently , i.e. , which requires the minimum user scaling . for the _ target dof _ per transmitter @xmath92 , we find the optimal target dof achieving strategy @xmath27 satisfying @xmath93 and derive the required user scaling . note that the definition of dof in this paper is extended from the conventional definition of dof in order to properly capture multiuser diversity gain in terms of achievable rate . achievable dof defined in depends on increasing speeds of @xmath13 and @xmath24 ; and can have non - zero values even larger than one if the number of users properly scales with the transmit power . from the definitions of the rate gain term and the rate loss term given in and , respectively , the strategies which achieve the target dof @xmath16 are given by @xmath94 the following lemma shows that we do not need to consider all of the candidate strategies in but take into account only a subset of to find the optimal target dof achieving strategy . [ lemma : dof_terms_range ] for any non - negative target dof , the optimal dof achieving strategy is in the set @xmath95,~d_1-d_2=d \big\}. \label{eqn : dof_term_range}\end{aligned}\ ] ] at each channel realization , the achievable dof has the form of @xmath96 where @xmath97 and @xmath98 are its own signal power and the interfering signal power at the selected user , respectively . since the function is an increasing function of @xmath99 and a decreasing function of @xmath100 , for an increase of , the multiuser dimension should be used for increasing @xmath99 , for decreasing @xmath100 , or mixture of them . this fact results in . lemma 1 provides a basic guideline of using the multiuser dimension ; multiuser diversity should not be used for either decreasing dof gain term or increasing dof loss term . since the optimal target dof achieving strategy is obtained in the reduced set of candidate strategies , we consider the dof gain term larger than one and dof loss term smaller than one , i.e. , @xmath101 and @xmath102 $ ] , in the latter parts of this paper . in our system model , each user suffers from @xmath1 interfering channels which is larger than or equal to the number of receive antennas , i.e. , @xmath103 . since the interfering channels are isotropic and independent of each other , they span @xmath14-dimensional space . thus , the whole signal space at the receiver is corrupted by interfering signals , and hence the dof loss term becomes one if no effort is made to align interfering signals . on the other hand , the dof loss can be reduced by aligning interfering signals in smaller dimensional subspace . for example , if the interfering signals are perfectly aligned in @xmath104-dimensional subspace , they can be nullified by postprocessing so that we can make the dof loss zero . the transmitter can exploit the multiuser dimensions to align interfering signals by simply selecting a user whose interfering channels are most aligned . thus , each user needs to measure how much the interfering channels are aligned in @xmath104-dimensional subspace at the receiver . we call this measure as the _ interference alignment measure_. in this section , we geometrically interpret the interfering channels and define the interference alignment measure at each user . the interference alignment measure will be used for computing the reducible dof loss via multiuser diversity in section iv . let @xmath105 be the surface of the @xmath14-dimensional unit hypersphere centered at the origin , i.e. , @xmath106 for an arbitrary unit vector @xmath107 and an arbitrary non - negative real number @xmath108 , we can divide @xmath105 into two parts , @xmath109 and @xmath110 , given by @xmath111 when @xmath112 , two parts @xmath113 and @xmath114 are represented in fig . [ fig : sphere_s1s2 ] . let @xmath115 be the surface area of @xmath116 for @xmath117 . the surface area of an @xmath14-dimensional complex unit hypersphere is given by @xmath118 , and it was shown that @xcite @xmath119 which is invariant with @xmath120 . therefore , we obtain @xmath121 from the relationship @xmath122 . from this fact , we obtain the following lemma . [ lemma : p_n_lambda ] let @xmath123 be independent and isotropic unit vectors in @xmath124 . for an arbitrary unit vector @xmath125 and @xmath126 $ ] , the probability that @xmath127 contains @xmath128 becomes @xmath129 = \left(1-(1-\lambda)^{n_r-1}\right)^m , \label{eqn : prob_m_lambda}\end{aligned}\ ] ] which is invariant with @xmath120 . from the ratio of @xmath130 and @xmath131 , we obtain @xmath132 = \frac{a(s_2(\mathbf{c } , \lambda))}{a(s_0 ) } = 1-(1-\lambda)^{n_r-1 } , \quad \forall i.\end{aligned}\ ] ] since @xmath123 are independent of each other , it is satisfied that @xmath133 & = \pr[\mathbf{g}_1 \in s_2 ( \mathbf{c } , \lambda)]^m , { \nonumber}\end{aligned}\ ] ] which is given in . in this subsection , we define the interference alignment measure at each user . the dof loss is determined by how much the interfering channels are closely aligned in @xmath104-dimensional subspace . only if interfering channels are perfectly aligned in @xmath104-dimensional subspace , we can have zero dof loss . the interference alignment measure is used for computing the dof loss at each user . let @xmath134 be the @xmath135 normalized interfering channels at a user and @xmath136 be the interference alignment measure among them . consider the following optimization problem : @xmath137.{\nonumber}\end{aligned}\ ] ] from the definition of @xmath138 given in , this problem is equivalent to @xmath139,{\nonumber}\end{aligned}\ ] ] which can be solved by linear programming @xcite . let @xmath140 be the solution of the above problem . then , @xmath141 has the smallest surface area among all @xmath142 containing @xmath134 . using @xmath143 , we can divide an @xmath14-dimensional space into two subspaces which are the one - dimensional subspace spanned by @xmath143 and the @xmath104-dimensional complementary subspace denoted by @xmath144 . if there exists @xmath143 such that @xmath145 , it is satisfied that @xmath146 and @xmath147 , and hence @xmath148 becomes zero . in this case , we can say that the interfering channels are perfectly aligned in @xmath104-dimensional subspace in @xmath124 . note that @xmath149 is an @xmath104-dimensional subspace orthogonal to @xmath150 , and @xmath151 is the @xmath14-dimensional complex hypersphere , @xmath105 . when @xmath148 is the smaller , the vectors are the more aligned in the @xmath104-dimensional subspace , @xmath144 . thus , we will use @xmath152 as an _ interference alignment measure _ to quantify how much the interfering channels are closely aligned in an @xmath104-dimensional subspace , i.e. , @xmath153).\label{eqn : lambda}\end{aligned}\ ] ] in other words , we use the mini - max distance of the interfering channels from an @xmath104-dimensional subspace . in fig . [ fig : iam_concept ] , the interference alignment measure is geometrically represented . the more the interfering channels are aligned , the smaller the interference alignment measure becomes . since the interference alignment measure is obtained from the optimization problem , the exact distribution is difficult to find . instead , we obtain the lower bound for the cumulative distribution function ( cdf ) of the interference alignment measure in the following lemma . [ lemma : prob_bar_d_m ] when @xmath154 , the probability that the interference alignment measure @xmath136 is smaller than @xmath155 $ ] is lower bounded on @xmath156 } \ge \left(1-(1-\lambda)^{n_r-1}\right)^{k - n_r}. \label{eqn : am1_prob}\end{aligned}\ ] ] we consider two events : @xmath157 where @xmath158 is the @xmath14-dimensional unit vector such that @xmath159 . by the definition of the interference alignment measure given in , @xmath160 is true whenever @xmath161 is true , equivalently , @xmath162 } \ge { \mathrm{pr}\left[\mathrm{(e2)}\right]}$ ] . the probability of @xmath161 is obtained by @xmath163 } & = { \mathrm{pr}\left[s_2(\bar{\mathbf{c } } , \lambda ) \supset \{\mathbf{g}_1 , \ldots , \mathbf{g}_{k-1}\}\right]}{\nonumber\\}&\stackrel{(a)}{= } { \mathrm{pr}\left[s_2(\bar{\mathbf{c } } , \lambda ) \supset \{\mathbf{g}_{n_r } , \ldots , \mathbf{g}_{k-1}\}\right]}{\nonumber\\}&\stackrel{(b)}{= } \left(1-(1-\lambda)^{n_r-1}\right)^{k - n_r},\end{aligned}\ ] ] where the equality @xmath164 is from the definition of @xmath158 such that @xmath165 . also , the equality @xmath166 holds from lemma [ lemma : p_n_lambda ] and from the fact that @xmath158 is independent of @xmath167 . thus , we obtain @xmath168 } \ge \left(1-(1-\lambda)^{n_r-1}\right)^{k - n_r}.\end{aligned}\ ] ] the remaining question is how much we can reduce the interference alignment measure via user selection . in the first user group , the @xmath53th user has @xmath1 interfering channels , @xmath169 . the interference alignment measure at the @xmath53th user can be written by @xmath170 where @xmath171 is the normalized interfering channel , i.e. , @xmath172 . thus , the achievable smallest interference alignment measure via user selection is given by @xmath173 obviously , the smallest interference alignment measure will decrease as the number of users increases . in the following lemma , we find the relationship between and the number of total users ( i.e. , @xmath13 ) . [ lemma : lambda_n1 ] when there are @xmath13 users , the expectation of the smallest interference alignment measure is upper bounded on @xmath174 < n^{-\frac{1}{k - n_r}}.\label{eqn : am1_bound}\end{aligned}\ ] ] the complementary cdf of is bounded on @xmath175 } & \qquad={\mathrm{pr}\left[\mathfrak{q } \big ( \tilde{\mathbf{h } } _ { n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k}\big ) \ge \lambda \textrm{~for all~ } n\right]}{\nonumber\\}&\qquad=\prod_{n=1}^n{\mathrm{pr}\left [ \mathfrak{q } \big ( \tilde{\mathbf{h } } _ { n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k}\big ) \ge \lambda\right]}{\nonumber\\}&\qquad=\left(1-{\mathrm{pr}\left [ \mathfrak{q } \big ( \tilde{\mathbf{h } } _ { n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k}\big ) \le \lambda\right]}\right)^n{\nonumber\\}&\qquad\stackrel{(a ) } { < } \big[1- ( 1-(1 - \lambda)^{n_r-1})^{k - n_r}\big]^n , \label{eqn : prob_bound}\end{aligned}\ ] ] where @xmath176 $ ] , and the inequality @xmath164 holds from lemma [ lemma : prob_bar_d_m ] . using this bound , we obtain as @xmath177 & = \int_0 ^ 1 { \mathrm{pr}\left[\min_n \mathfrak{q } \big ( \tilde{\mathbf{h } } _ { n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k}\big)\ge \lambda\right ] } d \lambda { \nonumber\\}&\le \int_0 ^ 1 \big[1- ( 1 - ( 1- \lambda ) ^{n_r-1})^{k - n_r}\big]^n d \lambda{\nonumber\\}&\stackrel{(a)}{\le } \int_0 ^ 1 \big[1- ( 1 - ( 1- \lambda))^{k - n_r}\big]^n d \lambda{\nonumber\\}&\stackrel{(b)}{= } \frac{1}{k - n_r}\beta\left(\frac{1}{k - n_r } , n+1\right){\nonumber\\}&\stackrel{(c)}{= } \frac{\gamma\left(1+\frac{1}{k - n_r}\right)\gamma(n+1 ) } { \gamma\left(n+ 1 + \frac{1}{k - n_r}\right ) } { \nonumber\\}&\stackrel{(d ) } { < } n^{-\frac{1}{k - n_r}},{\nonumber}\end{aligned}\ ] ] where the inequality @xmath164 is due to @xmath178 for @xmath108 , and the equality @xmath166 holds from the representation of beta function @xcite @xmath179 the equality @xmath180 comes from the definition of the beta function @xmath181 and the property of the gamma function @xmath182 . in the right - hand - side of the equality @xmath180 , it holds @xmath183 because @xmath184 for @xmath185 . also , it is satisfied that @xmath186 where @xmath187 is from the gautschi s inequality @xcite given by @xmath188 with @xmath189 and @xmath190 . thus , the inequality @xmath191 holds . in this section , we derive the optimal strategies of exploiting multiuser diversity for the target dof @xmath16 . we first decompose the target dof @xmath16 into the dof gain term @xmath20 and the dof loss term @xmath21 such that @xmath192 , and find the required user scalings for @xmath20 and @xmath21 , respectively . then , the optimal target dof achieving strategy is derived by determining the optimal combination @xmath193 which requires the minimum user scaling for the target dof @xmath16 . in this subsection , we find the required user scaling to reduce the dof loss . via user selection , the rate loss term given in can be minimized by @xmath194 . \label{eqn : rloss_min}\end{aligned}\ ] ] this value is upper bounded on @xmath194 & \stackrel{(a)}{= } \mathbb{e}_{\vert\mathbf{h}\vert , \tilde{\mathbf{h } } } \left[\min_{n,\mathbf{v}_n } ~\log_2\left(1 + p \sum_{k=2}^{k } \vert\mathbf{h}_{n , k}\vert^2\vert \mathbf{v}_{n}^\dagger \tilde{\mathbf{h}}_{n , k } \vert^2\right ) \right]{\nonumber\\}&\stackrel{(b)}{\le } \mathbb{e}_{\tilde{\mathbf{h } } } \left[\min_{n,\mathbf{v}_n } ~\mathbb{e}_{\vert\mathbf{h}\vert}\log_2\left(1 + p \sum_{k=2}^{k } \vert\mathbf{h}_{n , k}\vert^2\vert \mathbf{v}_{n}^\dagger \tilde{\mathbf{h}}_{n , k } \vert^2\right ) \right]{\nonumber\\}&\stackrel{(c)}{\le } \mathbb{e}_{\tilde{\mathbf{h } } } \left[\min_{n,\mathbf{v}_n } ~\log_2\left(1 + n_rp \sum_{k=2}^{k } \vert \mathbf{v}_{n}^\dagger \tilde{\mathbf{h}}_{n , k } \vert^2\right ) \right]{\nonumber\\}&\stackrel{(d)}{\le } \mathbb{e}_{\tilde{\mathbf{h } } } \left[\min_{n } ~\log_2\left(1 + n_rp ( k-1)\mathfrak{q}(\tilde{\mathbf{h}}_{n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k } ) \right)\right]{\nonumber\\}&\stackrel{(e)}{\le } ~\log_2\left(1 + n_rp ( k-1 ) \mathbb{e}_{\tilde{\mathbf{h}}}\left [ \min_{n } \mathfrak{q}(\tilde{\mathbf{h}}_{n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k } ) \right]\right){\nonumber\\}&\stackrel{(f)}{\le } ~\log_2\left(1 + n_rp ( k-1 ) n^{-\frac{1}{k - n_r}}\right ) , \label{eqn : rloss_bound}\end{aligned}\ ] ] where the equality @xmath164 is obtained by decomposing the channel vector into direction and magnitude independent of each other such that @xmath195 . the inequality @xmath166 holds because the minimum of the average is larger than the average of the minimum . the inequality @xmath180 is from the jensen s inequality and @xmath196 . also , the inequality @xmath191 holds from the fact that @xmath197 & \le \min_{\mathbf{v}_n } \left[(k-1 ) \max_{2\le k\le k } \vert \mathbf{v}_n^\dagger \tilde{\mathbf{h}}_{n , k}\vert^2 \right]{\nonumber\\}&= ( k-1 ) \mathfrak{q } ( \tilde{\mathbf{h}}_{n,2 } , \ldots , \tilde{\mathbf{h}}_{n , k}),\end{aligned}\ ] ] where @xmath198 is the interference alignment measure at the user @xmath53 given in . the inequality @xmath187 is from the jensen s inequality , and the inequality @xmath199 holds from lemma [ lemma : lambda_n1 ] . we obtain the following theorem . [ theorem : n_scaling_dof_loss ] we can obtain the dof loss term @xmath200 $ ] when the number of users in each group scales as @xmath201 to obtain the dof loss term @xmath21 , it is enough to make satisfying @xmath202 which is achieved if @xmath203 . a tighter upper bound of the rate loss term than could exist , but the derived upper bound in enables us to compare the increasing speeds of the transmit power and the required number of users , which is the crucial factor of dof calculation . the scaling law of the required number of users obtained from , which is derived in theorem [ theorem : n_scaling_dof_loss ] , is enough to find the optimal target dof achieving strategy as shown in section [ sec : dof_achieving_strategy ] . we also find the required user scaling to increase the dof gain term . from the definition of the rate gain term given in , the maximum rate gain term obtained by user selection is @xmath204 . \label{eqn : rgain_max}\end{aligned}\ ] ] this value is lower bounded on @xmath205 & \ge\mathbb{e}\left [ \max_{n , \mathbf{v}_n } \log_2\left(1 + p \vert \mathbf{v}_{n}^\dagger { \mathbf{h}_{n,1 } } \vert^2\right)\right]{\nonumber\\}&= \mathbb{e}\left [ \max_{n } \log_2\left(1 + p \vert { \mathbf{h}_{n,1 } } \vert^2\right)\right],\label{eqn : rgain_lower_bound}\end{aligned}\ ] ] and upper bounded on @xmath205 & \le \mathbb{e}\left [ \max_{n } \log_2\left(1 + p \sum_{k=1}^{k } \vert { \mathbf{h}_{n , k } } \vert^2\right)\right]{\nonumber\\}&\le \mathbb{e}\left [ \max_{n } \log_2\left(1 + pk \max_k \vert { \mathbf{h}_{n , k } } \vert^2\right)\right].\label{eqn : rgain_upper_bound}\end{aligned}\ ] ] since all @xmath206 are i.i.d . @xmath207 random variables , for sufficiently large @xmath13 , the bounds and acts like @xcite @xmath208 & \sim \log_2(1+p\log n ) { \nonumber\\}\mathbb{e}\left [ \max_{n , k } \log_2\left(1 + pk \vert { \mathbf{h}_{n , k } } \vert^2\right)\right ] & \sim \log_2(1+pk\log ( kn)).{\nonumber}\end{aligned}\ ] ] thus , when both @xmath13 and @xmath24 are large enough , act like @xmath209 , i.e , @xmath210 \sim \log_2(p\log n ) . \label{eqn : rgain_asymptotic}\end{aligned}\ ] ] therefore , we establish the following theorem . [ theorem : n_scaling_dof_gain ] the dof gain term @xmath211 is achievable when the number of users in each group scales as @xmath212 we use . by setting @xmath213 we obtain the required user scaling for the dof gain term @xmath20 given by @xmath25 . in theorem [ theorem : n_scaling_dof_loss ] and theorem [ theorem : n_scaling_dof_gain ] , we found the required user scalings for the dof loss term @xmath21 and the dof gain term @xmath20 , respectively . in this subsection , we find the optimal target dof achieving strategy which requires the minimum user scaling . we start with the following theorem . [ theorem : scaling1 ] for the target dof up to one , the whole multiuser dimensions should be devoted to minimizing the dof loss caused by interfering signals . the optimal dof achieving strategy for the target dof @xmath214 $ ] is @xmath215 , and the corresponding sufficient user scaling is @xmath216 in theorem [ theorem : n_scaling_dof_loss ] , we have shown that the target dof @xmath217 $ ] is achievable by reducing the dof loss term with the user scaling @xmath218 . on the other hand , this user scaling can not increase the dof gain term . substituting @xmath218 into the dof gain term @xmath20 becomes @xmath219 which is the same as when there is a fixed number of users as described in . that is , any other combinations @xmath220 which achieve the target dof @xmath16 requires larger user scaling than @xmath218 , where @xmath221 since @xmath222 . therefore , the optimal target dof achieving strategy is given by @xmath215 , and the sufficient user scaling is @xmath218 . now , we derive the target dof achieving strategy when the target dof @xmath16 is greater than one . to find the optimal dof achieving strategy , we firstly find the sufficient user scaling for an arbitrary strategy @xmath90 achieving dof @xmath223 . then , we show that the optimal target dof achieving strategy for the target dof @xmath224 is @xmath225 . [ lemma : general_scaling_d>1 ] for the target dof @xmath226 , the sufficient user scaling for an arbitrary strategy @xmath90 achieving dof @xmath227 is given by @xmath228 where @xmath222 and @xmath102 $ ] . as a target dof achieving scheme , we consider a two - stage user selection scheme ; the first stage is to increase the dof gain term , and the second stage is to decrease the dof loss term . the considered two stage user selection strategy is illustrated in fig . [ fig : user_selection_2stage ] . we randomly divide total @xmath13 users into @xmath229 subgroups having @xmath230 users each such that @xmath231 . then , the user selection in each stage is performed as follows . stage 1 : in each subgroup , a single user having the largest channel gain is selected among @xmath230 users . as a result , we have @xmath229 selected users after stage 1 . * stage 2 : among the @xmath229 users , the transmitter selects a single user to minimize the dof loss term . in stage 1 , the dof gain term @xmath20 is obtained at each selected user when @xmath232 as stated in theorem [ theorem : n_scaling_dof_gain ] . in stage 2 , we can make the dof loss term @xmath21 when @xmath233 as shown in theorem [ theorem : n_scaling_dof_loss ] . thus , the target dof @xmath226 with the strategy @xmath90 such that @xmath192 can be obtained by the user scaling @xmath234 , which is given in . from lemma [ lemma : general_scaling_d>1 ] , we obtain the optimal dof achieving strategy for the target dof @xmath235 in following theorem . [ theorem : scaling2 ] the optimal target dof achieving strategy for @xmath236 is to increase the dof gain term to @xmath16 and to perfectly eliminate the dof loss , i.e. , @xmath237 . consequently , the sufficient user scaling for target dof @xmath226 becomes @xmath238 the proof is similar to that of theorem [ theorem : scaling1 ] . from lemma [ lemma : general_scaling_d>1 ] , we can obtain the target dof @xmath226 by the strategy @xmath239 with the sufficient user scaling given in . however , this scaling can not increase the dof gain term larger than @xmath16 even when the user scaling is only used to increase the dof gain term . substituting into , we still have @xmath240 this implicates that the user scaling given in is sufficient for the strategy @xmath239 but not enough for other strategies @xmath241 as well as @xmath242 which requires the higher user scaling than that of @xmath241 , where @xmath243 $ ] . therefore , the optimal strategy for the target dof @xmath226 becomes @xmath244 . in fig . [ fig : dof_achieving_strategy ] , the optimal dof achieving strategy @xmath27 is plotted according to the target dof @xmath23 . in this section , we discuss how the optimal target dof achieving strategy can be realized by practical user selection schemes . the practical schemes considered in this section require no cooperation and no information exchange among the transmitters . for practical scenarios , we assume that each user has knowledge of channel state information ( csi ) of the direct channel and the covariance matrix of the received signal without explicit knowledge of csi of the interfering channels . that is , the @xmath53th user knows csi of its own desired channel @xmath245 and the covariance matrix of the received signal @xmath246 = \mathbf{i}_{n_r } + p\sum_{k=1}^k \mathbf{h}_{n , k } \mathbf{h}_{n , k } ^\dagger$ ] . from these values , the user @xmath53 easily obtains the interference covariance matrix denoted by @xmath247 such as @xmath248 - p\mathbf{h}_{n,1}\mathbf{h}_{n,1}^\dagger -\mathbf{i}_{n_r}.\end{aligned}\ ] ] therefore , the achievable rate at the first transmitter given in can be rewritten by @xmath249 to increase @xmath73 , various user selection schemes can be considered , but we focus on several popular techniques in the following subsections to maximize the postprocessed snr @xmath250 , to minimize the postprocessed inr @xmath251 , and to maximize the postprocessed sinr @xmath252 . in the max - snr user selection scheme , each user maximizes the postprocessed snr , and the transmitter selects the user having the maximum postprocessed snr . consequently , the postprocessed snr at the selected user becomes @xmath253 \stackrel{(a)}{= } \max_n p\vert { \mathbf{h}_{n,1}}\vert^2,\end{aligned}\ ] ] where the equality @xmath164 holds when the @xmath53th user adopts the postprocessing vector @xmath254 . thus , the selected user denoted by @xmath255 becomes @xmath256 and the desired channel gain at each user ( @xmath257 for the user @xmath53 ) should be informed to the transmitter . in the min - inr user selection scheme , each user minimizes the postprocessed inr , and the transmitter selects the user having the minimum postprocessed inr . thus , the postprocessed inr at the selected user becomes @xmath263 \stackrel{(a)}{= } \min_n \left[\lambda _ { \min } \left(\mathbf{r}_{n}\right)\right],\end{aligned}\ ] ] where the equality @xmath164 is obtained by the postprocessing vector of the @xmath53th user @xmath264 the required feedback information from the @xmath53th user is @xmath265 , and index of the selected user denoted by @xmath266 becomes @xmath267.\end{aligned}\ ] ] note that this scheme minimizes the rate loss term defined in . using the min - inr scheme , the transmitter can decrease the dof loss term while the dof gain term remains to be one . therefore , the min - inr scheme realizes the optimal target dof achieving strategy @xmath268 for the target dof @xmath30 $ ] . the required number of users by the min - inr scheme for the target dof @xmath30 $ ] scales like @xmath269 which is the required user scaling of the optimal target dof achieving strategy when the target dof is @xmath270 $ ] as shown in theorem [ theorem : scaling1 ] . the max - sinr user selection scheme is known to maximize the achievable rate at the transmitter although it requires additional complexity for postprocessing at the receivers . the achievable rate by the max - sinr scheme denoted by @xmath271 becomes @xmath272 . \label{eqn : r_sinr}\end{aligned}\ ] ] at each channel realization , the postprocessed sinr at the selected user is given by @xmath273 . \label{eqn : max_sinr}\end{aligned}\ ] ] to maximize the postprocessed sinr , the @xmath53th user adopts the postprocessing vector given by @xmath274 which is identical with the mmse - irc in @xcite . the corresponding postprocessed sinr at user @xmath53 becomes @xmath275 @xcite , and hence the selected user at the transmitter denoted by @xmath276 is given by @xmath277 [ lemma : scaling1 ] to obtain the target dof @xmath30 $ ] , the required user scaling of the max - sinr scheme is exactly the same as that of the min - inr scheme . from the fact that @xmath278 we obtain @xmath279 where the equality @xmath164 is because @xmath280 as shown in the proof of theorem [ theorem : scaling1 ] . therefore , the required user scaling for @xmath281 is exactly the same as the required user scaling for @xmath282 , equivalently , for @xmath283 . lemma [ lemma : scaling1 ] indicates that the max - sinr scheme realizes the optimal dof achieving strategy @xmath284 for the target dof @xmath30 $ ] . [ lemma : scaling2 ] the max - sinr scheme realizes the dof achieving strategy @xmath285 whenever the target dof @xmath16 is greater than 1 . since the max - sinr scheme is the optimal user selection scheme , it achieves dof @xmath226 with the user scaling @xmath286 as stated in theorem [ theorem : scaling2 ] . from the definition of , we obtain @xmath287 , and hence we have @xmath288 . as shown in , the sufficient user scaling for the max - sinr scheme to obtain the target dof @xmath16 can not increase the dof gain term larger than @xmath16 even if the whole user scaling is only devoted to increasing the dof gain term . this implicates that when we obtain the target dof @xmath226 by the max - sinr scheme with the user scaling @xmath286 , we obtain @xmath289 and have the dof gain @xmath16 at most ( i.e. , @xmath290 ) . therefore , the max - sinr scheme can only have @xmath291 if @xmath292 . for the target dof @xmath226 , the two - stage user selection scheme described in the proof of lemma [ lemma : general_scaling_d>1 ] can be adopted . more specifically , the transmitter selects the users by the max - snr scheme in the first stage . then , in the second stage , the transmitter selects a single user by the min - inr scheme or the max - sinr scheme . as shown in the proof of lemma [ lemma : general_scaling_d>1 ] , the two - stage user selection scheme can realize the optimal target dof achieving strategy for the target dof @xmath226 . in this section , we extend our system to interfering mimo bc cases . more specifically , each transmitter with @xmath32 antennas sends @xmath32 independent streams over @xmath32 orthonormal random beams using equal power allocation . similar to the user selection procedure in @xcite , each transmitter broadcasts @xmath32 orthonormal random beams , and each user feeds @xmath32 scalar values corresponding to all beams back to the transmitter . the feedback information corresponding to each stream such as snr , inr , and sinr can be easily found in a similar way to the simo case . a single user is selected for each beam , but the same user can be selected for different beams . however , it rarely occurs that multiple streams are transmitted for a single user as the number of users increases . when multiple streams are transmitted for a single user , the user is assumed to decode each stream treating the other streams as interferences . we denote the orthonormal random beams by @xmath293 which satisfies that @xmath294 and @xmath295 for all @xmath296 . we start from the following remark . let @xmath297 be the channel matrix whose elements are i.i.d . gaussian random variables . then , for an arbitrary unitary matrix @xmath298 ( i.e. , @xmath299 ) , the distributions of @xmath300 and @xmath301 are identical . since the @xmath14 columns of @xmath300 are independent and isotropic random vectors in @xmath124 , so are the @xmath14 columns of @xmath302 . owing to remark 1 , the @xmath0-transmitter mimo interfering bc is statistically identical with the @xmath303-transmitter simo interfering bc . let @xmath304 be the channel matrix from the @xmath62th transmitter to the @xmath53th user in the first user group . since the random beams satisfy that @xmath305^\dagger [ \mathbf{u}_1 , \ldots , \mathbf{u}_{n_t } ] = [ \mathbf{u}_1 , \ldots , \mathbf{u}_{n_t } ] [ \mathbf{u}_1 , \ldots , \mathbf{u}_{n_t}]^\dagger = \mathbf{i}_{n_t}$ ] , the user @xmath53 in the first user group has @xmath303 independent and isotropic channel vectors @xmath306 ~~ i\in[n_t],{\nonumber}\end{aligned}\ ] ] formed by the random beams and channel matrices from all transmitters . if the @xmath53th user in the first group is served by the @xmath43th random beam , the user has desired channel @xmath307 and the @xmath308 interfering channels , which correspond to @xmath309 inter - stream interfering channels @xmath310 and @xmath311 inter - transmitter interfering channels @xmath312 } \{\mathbf{h}_{n , k}\mathbf{u}_j\}_{k=2}^k.{\nonumber}\end{aligned}\ ] ] consequently , each random beam can be regarded as a single antenna transmitter with the transmit power @xmath313 . this fact leads to the following theorems . [ theorem : mimo_bc ] in a mimo bc where a transmitter with @xmath32 antennas supports @xmath32 users among @xmath13 users with @xmath314 antennas each , the optimal dof achieving strategy for the target dof @xmath315)$ ] is @xmath316 and requires the number of users to scale as @xmath317 . for the target dof @xmath318 , the optimal dof achieving strategy is @xmath239 and requires the number of users to scale as @xmath319 . the dof gain term @xmath320 is obtained when each stream achieves dof gain term @xmath321 . thus , with the same procedure given in section [ sec : dof_gain_scaling ] , we can easily show that the dof gain term @xmath320 is obtained when @xmath322 . on the other hand , the dof loss term @xmath323 is obtained when each stream achieves dof loss term per stream @xmath324 . as stated earlier , using @xmath32 orthonormal random beams at the transmitter each of which independently supports a single user , the mimo bc can be translated into an @xmath32-transmitter simo ic where each transmitter supports one of @xmath13 users with @xmath14 antennas . in this case , dof loss term @xmath323 , i.e. , dof loss @xmath325 per stream , is obtained when @xmath326 . therefore , we can conclude that the optimal dof achieving strategy for the target dof @xmath315)$ ] is @xmath327 and requires the number of users to scale as @xmath328 . also , for the target dof @xmath318 , the optimal dof achieving strategy is @xmath239 and requires the number of users to scale as @xmath329 . [ theorem : mimo_ibc ] consider a @xmath0-transmitter interfering mimo bc where the @xmath62th transmitter with @xmath330 antennas supports @xmath330 users among @xmath331 users with @xmath332 antennas each . at the @xmath62th transmitter , the optimal dof achieving strategy for the target dof @xmath333)$ ] is @xmath334 and requires the number of users to scale as @xmath335 . for the target dof @xmath37 , the optimal dof achieving strategy becomes @xmath239 and requires the number of users to scale as @xmath336 . the proof is similar to that of theorem [ theorem : mimo_bc ] . the @xmath62th transmitter obtains dof gain term @xmath337 when each stream obtains dof gain term @xmath338 , and the required user scaling is exactly given by @xmath339 . on the other hand , the @xmath62th transmitter obtains @xmath340 when the dof loss term per stream becomes @xmath341 . using @xmath330 orthonormal random beams at each transmitter each of which independently supports a single user , the interfering mimo bc can be translated into an @xmath342-transmitter simo ic where each transmitter supports a single user among @xmath331 users with @xmath343 antennas . thus , the @xmath62th transmitter obtains the dof loss term @xmath340 when @xmath344 . therefore , we can conclude that the optimal dof achieving strategy of the @xmath62th transmitter for the target dof @xmath333)$ ] is @xmath334 and obtained when the number of users scales as @xmath345 . also , for the target dof @xmath346 , the optimal dof achieving strategy is @xmath239 and requires the number of users to scale as @xmath336 . in this section , we first compare achievable rates of the practical user selection schemes for given number of users . then , we check if the target dof can be achievable with increasing number of users by showing achievable rates per transmitter for the practical user selection schemes . we have also considered two time division multiple access ( tdma ) schemes . in the first tdma scheme ( tdma1 ) , a single transmitter operates at each time so that @xmath2 dof is achieved at each transmitter . in the second tdma scheme ( tdma2 ) , only @xmath14 of @xmath0 transmitters operate at each time so that @xmath347 dof is achieved at each transmitter . [ fig : oia_k4_a10_n10 ] shows the achievable rates of each transmitter for various user selection schemes in ibc when there are 4 transmitters and each transmitter has 10 users with three receive antennas each . it is confirmed that the achievable rates are saturated in the high snr region and the achievable dof per transmitter becomes zero for the fixed number of users . now , we show that the target dof can be achievable if the number of users properly scales . in fig . [ fig : oia_k4_a10_kn ] , the number of users scales as @xmath348 , i.e. , @xmath349 for the target dof one . specifically , two user scaling @xmath350 and @xmath351 are considered , and other configurations except the number of users are the same as those in fig . [ fig : oia_k4_a10_n10 ] . fig . [ fig : oia_k4_a10_kn ] verifies that the min - inr and the max - sinr schemes achieve dof one per transmitter as predicted in theorem [ theorem : scaling1 ] and lemma [ lemma : scaling1 ] . in fig . [ fig : oia_k4_a10_n_powerp ] , we consider two different user scaling @xmath352 and @xmath353 from those in fig . [ fig : oia_k4_a10_kn ] . according to theorem [ theorem : scaling1 ] and lemma [ lemma : scaling1 ] , the achievable dof at each transmitter by either the max - sinr scheme or the min - inr scheme is @xmath16 when the number of users scales as @xmath354 . as predicted , fig . [ fig : oia_k4_a10_n_powerp ] shows that the achieved dof per transmitter is 0.5 and 1 when @xmath352 and @xmath353 , respectively , by either the min - inr scheme or the max - sinr scheme . we first studied the optimal way of exploiting multiuser diversity in the @xmath0-transmitter simo ibc where each transmitter with a single antenna selects a user and the number of transmitters is larger than the number of receive antennas at each user . we proved that the multiuser dimensions should be used first for decreasing the dof loss caused by interfering signals ; the whole multiuser dimensions should be exploited to reduce the dof loss term to @xmath355 for the target dof @xmath356 $ ] , while the multiuser dimensions should be devoted to making the dof loss zero and then to increasing the dof gain term to @xmath16 for the target dof @xmath357 . we also derived the sufficient user scaling for the target dof . the dof per transmitter @xmath33 $ ] is obtained when the number of users scales as @xmath358 , and the dof per transmitter @xmath357 is achieved when the number of users scales as @xmath359 . also , we extended the results to the @xmath0-transmitter mimo ibc where each transmitter having the multiple antennas supports the multiple users . k. gomadam , v. r. cadambe , s. a. jafar , `` approaching the capacity of wireless networks through distributed interference alignment , '' in _ proc . of ieee global telecommunications conference _ , pp . 16 , dec . 2008 . y. huang and b. rao , random beamforming with heterogeneous users and selective feedback : individual sum rate and individual scaling laws " , _ ieee trans . wireless commun . 5 , pp . 20802090 , may 2013 . j. h. lee and w. choi , opportunistic interference aligned user selection in multi - user mimo interference channels , " in _ proc . of ieee global telecommunications conference _ , miami , fl , usa , dec . j. h. lee and w. choi , interference alignment by opportunistic user selection in 3-user mimo interference channels , " in _ proc . of ieee international conference on communications _ , kyoto , japan , june , 2011 . j. h. lee and w. choi , on the achievable dof and user scaling law of opportunistic interference alignment in 3-transmitter mimo interference channels , " _ ieee trans . wireless commun . 12 , no . 6 , pp . 2743 - 2753 , june 2013 . j. h. lee and w. choi , opportunistic interference alignment by receiver selection in a k - user 1 x 3 simo interference channel , " in _ proc . of ieee global telecommunications conference _ , houston , tx , usa , dec . 2011 . t. kim , d. j. love , and b. clerckx , instantaneous degrees of freedom of downlink interference channels with multiuser diversity , " in _ proc . of asilomar conference on signals , systems , and computers _ , pp.12931297 , nov . b. c. jung , d. park , and w .- y . shin , opportunistic interference mitigation achieves optimal degrees - of - freedom in wireless multi - cell uplink networks , " _ ieee trans . 60 , no . 7 , july 2012 . j. joung and a. h. sayed , user selection methods for multiuser two - way relay communications using space division multiple access , " _ ieee trans . wireless commun . _ , vol . 9 , no . 7 , pp 21302136 , july 2010 . y. ohwatari , n. miki , t. asai , t. abe , and h. taoka , performance of advanced receiver employing interference rejection combining to suppress inter - cell interference in lte - advanced downlink " , in _ proc . ieee vehicular technology conference ( vtc - fall ) _ , sep . y. tokgoz , b. d. rao , m. wengler , and b. judson , performance analysis of optimum combining in antenna array systems with multiple interferers in flat rayleigh fading , " _ ieee trans . 10471050 , july 2004 .	this paper investigates how multiuser dimensions can effectively be exploited for target degrees of freedom ( dof ) in interfering broadcast channels ( ibc ) consisting of @xmath0-transmitters and their user groups . first , each transmitter is assumed to have a single antenna and serve a singe user in its user group where each user has receive antennas less than @xmath0 . in this case , a @xmath0-transmitter single - input multiple - output ( simo ) interference channel ( ic ) is constituted after user selection . without help of multiuser diversity , @xmath1 interfering signals can not be perfectly removed at each user since the number of receive antennas is smaller than or equal to the number of interferers . only with proper user selection , non - zero dof per transmitter is achievable as the number of users increases . through geometric interpretation of interfering channels , we show that the multiuser dimensions have to be used first for reducing the dof loss caused by the interfering signals , and then have to be used for increasing the dof gain from its own signal . the sufficient number of users for the target dof is derived . we also discuss how the optimal strategy of exploiting multiuser diversity can be realized by practical user selection schemes . finally , the single transmit antenna case is extended to the multiple - input multiple - output ( mimo ) ibc where each transmitter with multiple antennas serves multiple users . multiuser diversity , degrees of freedom , interference alignment measure , interfering broadcast channel
the investigation of quantum phase transitions in strongly correlated electronic systems has been an active field of research during the last two decades.@xcite recently , there has been growing interest in so - called impurity quantum phase transitions these correspond to qualitative changes in the ground - state properties of discrete degrees of freedom ( a spin @xmath1 in the simplest case ) upon changing their couplings to an external environment . on the one hand , impurity quantum phase transitions can be realized , e.g. , in mesoscopic systems , such as quantum dot devices , which provide the high tunability required to access the transition point . on the other hand , the study of impurity problems is motivated through the dynamical mean - field theory@xcite ( dmft ) , which maps lattice models of strongly correlated electrons assuming a momentum - independent self energy onto effective self - consistent impurity models . in this language , certain bulk quantum phase transition correspond to impurity transitions in anderson or kondo - type models . impurity quantum phase transition occur in different contexts and setups . one possibility is a single magnetic impurity coupled to an unconventional fermionic host . a paradigmatic example being the so - called pseudogap kondo model,@xcite originally motivated by the study of magnetic impurities in @xmath2-wave superconductors , where a spin @xmath1 is coupled to the spin density of fermionic quasiparticles with a power - law density of states ( dos ) , @xmath3 . the lack of low - energy states ( as compared to the standard kondo model@xcite ) causes a competition between the local - moment dynamics and the formation of a strong - coupling ground state and leads to rich quantum critical behavior.@xcite other fermionic impurity models display quantum phase transitions due to the competition between differently screened states , with the two - channel kondo@xcite and the two - impurity kondo models@xcite being popular examples . the majority of these models were recently reviewed in ref . . quantum impurities with a _ bosonic _ environment constitute a distinct class of models which can feature quantum phase transitions as well . a well - studied model is the spin - boson model , originally introduced to capture dissipation effects at the quantum level . here , a generic two - level system ( dubbed spin ) is subject to the competing influence of a transverse field ( causing tunneling between the two levels ) and a longitudinally - coupled set of harmonic oscillators ( providing friction ) . in the case of ohmic dissipation , this results in a kosterlitz - thouless phase transition at zero temperature , tuned by a variation of the dissipation strength.@xcite note that the ohmic spin - boson model is equivalent to a kondo model with anisotropic spin exchange terms , and hence does not constitute a distinct universality class . ( this is different for sub - ohmic damping , see refs . . ) the topic of impurities in quantum magnets has motivated studies of the so - called bose - kondo model ( which may also be viewed as a spin - isotropic spin - boson model with zero transverse field ) this model has a a stable intermediate - coupling fixed point and no screening;@xcite this will be reviewed briefly in sec . [ vojtabura ] . we note that the above - mentioned bosonic impurity models can be extended to include a fermionic environment as well , resulting in bose - fermi kondo hamiltonians , where the interactions of the impurity with the bosonic and fermionic baths compete.@xcite models of this type have been proposed to describe quantum criticality in heavy - fermion compounds within extended dmft,@xcite and could as well be realized in certain quantum - dot setups.@xcite we have recently introduced@xcite a distinct class of su(2)-symmetric bosonic impurity models , where the environment consists of spin-1/2 objects , i.e. , spinons constituting the elementary excitations of certain frustrated quantum magnets . in contrast to the spin - boson type models mentioned above , where the environment always plays the role of a fluctuating ( classical ) magnetic field , our models display true quantum dynamics and allow singlet formation accompanied by complete screening . the purpose of this paper is to give a comprehensive renormalization group ( rg ) analysis of various versions of these bosonic kondo models , thus providing details which were left out in the short report of ref . : ( i ) the first example concerns the elementary bosonic excitations ( spinons ) in the z@xmath0 spin - liquid state of a triangular heisenberg model.@xcite . this model will be shown shown to display an impurity quantum phase transition , bearing some resemblance to the fermionic pseudogap kondo model . ( ii ) a second example that follows the same philosophy concerns the setup of a `` magnetic impurity '' embedded in an optical lattice filled with a multicomponent bosonic gas.@xcite here , the bosons are not collective modes originating from an underlying magnetic state , but are rather real bosonic atoms . in both cases , we derive the relevant quantum - field theory and study it using rg combined with epsilon expansion techniques . although the structure of the theories is reminiscent of fermionic kondo models , there are important differences : we demonstrate the importance of potential - scattering effects at the impurity location as well as the influence of bulk interactions . note that this paper will only study the rg flow at weak coupling ; for situations with runaway flow to strong coupling indicative of true screening different techniques need to be applied . we have demonstrated this in ref . , using strong - coupling toy models and large-@xmath4 techniques ; this will not be repeated here . furthermore , we restrict our attention to situations where the bulk state does not break a local symmetry , i.e. we focus on paramagnetic phases including the quantum critical points . the plan of the paper is as follows . in sec . [ orderparameters ] we summarize the order - parameter description of nearly - critical antiferromagnets , distinguishing collinear and non - collinear ordering . in the non - collinear case of interest to us , deconfined bosonic spinons@xcite emerge which correspond to fractional spin @xmath1 excitations in a z@xmath0 spin liquid phase.@xcite in sec . [ vojtabura ] we review the results of earlier work@xcite on a magnetic moment in a collinear antiferromagnet , described by a conventional @xmath5 theory with triplon excitations , leading to a spin - isotropic spin - boson - like model without screening . sec . [ rg ] is devoted to the central topic of this paper , namely the physics of a magnetic impurity in a z@xmath0 spin liquid . we present a microscopic derivation of the impurity action , which we analyze by rg methods , extending our earlier calculation@xcite to two - loop level . furthermore we propose a way to properly incorporate the bulk interactions using a mapping to an effective bosonic anderson impurity model . finally , in sec . [ canonical ] we study how an impurity two - level system couples to an atomic gas of canonical bosons carrying an internal degree of freedom ( pseudospin ) . again , we determine the phase diagram and critical properties using rg . a brief outlook will close the paper . details of the rg calculations are deferred to the appendices . there , we also point out that the rg flow derived in sec . [ canonical ] is identical to the one of a fully asymmetric fermionic kondo model , which may arise in the dmft context . different types of magnetic mott insulators naturally call for different types of order - parameter theories to describe the physics near the transition towards the symmetry - broken phase . in general , introducing an order parameter @xmath6 for the magnetic fluctuations near the ordering wavevector @xmath7 , a gradient expansion of the action naturally leads to a standard multicomponent @xmath5 theory ( or a non - linear sigma model ) . however , care is required for non - collinear spin correlations , as shown below . assuming a magnetic order characterized by a single ordering wave vector @xmath8 , the local spin operators can be parametrized as @xmath9 where @xmath10 denotes the position of lattice site @xmath11 , and @xmath12 is a complex vector field in spin space which describes the order - parameter direction and varies slowly in space and time . for a state with static collinear order we have fixed vectors @xmath13 with @xmath14 ; the @xmath15 in eq . on all sites @xmath11 point parallel or antiparallel w.r.t . a common axis . the undoped insulator la@xmath0cuo@xmath16 is of this type ( if we neglect the small spin canting due to dzyaloshinski - moriya interactions ) , with in - plane ordering wave - vector @xmath17 ( the lattice spacing is unity here and in the following ) . for this specific value of @xmath8 on a square lattice the order parameter is independent of @xmath18 . near quantum criticality , one is lead to the usual o(3 ) quantum @xmath5 theory in imaginary time : @xmath19 \,.\ ] ] for @xmath17 the vector field @xmath20 is now _ real _ ; for commensurate @xmath8 , @xmath20 is complex , but the phase degree of freedom is gapped . this theory has been studied at length in the literature,@xcite and its critical properties are relevant for the quantum phase transition from a nel state to a gapped paramagnet in several classes of unfrustrated antiferromagnets such as the bilayer square - lattice heisenberg model and coupled spin - ladder models . for general incommensurate @xmath8 , one has to parametrize @xmath21 , with a real vector @xmath22 and a phase field @xmath23 . interesting critical behavior may obtain here@xcite , which is not in the focus of this paper . non - collinear order is present if @xmath24 in ; the order may be coplanar or fully three - dimensional . the former situation is e.g. realized in triangular quantum antiferromagnets , where the @xmath15 now lie in a plane in spin space , rather than along a single axis . the simplest case of a non - collinear ordered state is the one where @xmath25 corresponding to a spin spiral . an elegant way to resolve the constraints is via a parametrization of the order parameter according to@xcite @xmath26 with the single constraint @xmath27 ( here @xmath28 ) , @xmath29 is the vector of pauli matrices , and @xmath30 is the fully antisymmetric matrix . the objects @xmath31 represent complex numbers ( and hence are not canonical bosons ) , they describe a mapping from the four - dimensional unit sphere s@xmath32 to s@xmath0 . obviously , this representation is unique up to a local z@xmath0 transformation @xmath33 with @xmath34 . thus , in the non - collinear case described by ( [ spiral],[couplingform ] ) , the long - distance fluctuations near a magnetic quantum critical point are described by a quantum field theory in terms of bosonic spinons:@xcite @xmath35 where the integration over the auxiliary field @xmath36 enforces the constraint on @xmath37 . in eq . an additional z@xmath0 gauge field , corresponding to the above - mentioned gauge redundancy , has been omitted this is justified provided that the gauge field is in a deconfined phase with massive fluctuations , which is the case for the triangular antiferromagnet . note that the paramagnetic phase of is a z@xmath0 spin liquid with gapped , deconfined bosonic spinons.@xcite let us note here that the fluctuations in a _ magnet may also be written in terms of bosonic spinons , leading to a schwinger boson representation:@xcite @xmath38 with the constraint @xmath39 . note the difference with eq . : in contrast to the non - collinear case , the accompanying gauge redundancy here is of u(1 ) symmetry . in such a situation , the gauge - field fluctuations are massless and can not be ignored in a proper description of the bulk magnetism . the associated compact gauge theory is in general confining,@xcite implying that bosonic spinons are bound in pairs and leading to the standard @xmath5 theory for triplon fluctuations.@xcite u(1 ) spin liquids with deconfined spinons may nevertheless arise : a special situation here are so - called deconfined critical points,@xcite where a description in terms of bosonic spinons coupled to a non - compact u(1 ) gauge field applies ( which implies the suppression of hedgehogs@xcite ) . impurities characterized by a gauge charge and embedded in a u(1 ) spin liquid have been studied in ref . ; the interplay with a magnetic impurity moment would be interesting to study , but is beyond the scope of the present paper . magnetic impurities embedded in a nearly critical collinear quantum antiferromagnet were analyzed in detail in refs . . the bulk action @xmath40 ( [ scoll ] ) , describing the physics close to the transition between the ordered nel state and the disordered spin - gap state , is supplemented by the coupling to an impurity quantum spin @xmath1 @xmath41 , leading to : @xmath42\,,\nonumber\\ \mathcal{s}_{\textrm{imp}}&= & \gamma_0 \int d \tau \vec s(\tau ) \cdot \vec\phi ( \vec{x}=0,\tau ) \label{fluctfield}\end{aligned}\ ] ] where @xmath43 is the real order - parameter field of ( [ scoll ] ) , and @xmath44 $ ] encodes the dynamics of the impurity spin @xmath41 . a weak - coupling rg analysis@xcite leads to the rg beta function for the dimensionless impurity coupling @xmath45 : @xmath46 where @xmath47 , @xmath48 is a renormalization scale , and the sign of the beta function is such that negative terms are rg relevant in the infrared . in eq . ( [ betagamma ] ) , @xmath49 denotes the renormalized bulk coupling in @xmath50 ( [ scoll ] ) which flows as in the standard @xmath5 theory.@xcite from eq . ( [ betagamma ] ) one finds a fixed - point value of the impurity coupling as @xmath51 . the critical fluctuations of the bulk plus impurity problem are thus controlled by a fixed point with couplings @xmath52 ( bulk ) and @xmath53 ( impurity ) , which are both of order @xmath54 , i.e. , perturbatively accessible near @xmath55 bulk dimensions . most importantly , the flow of @xmath45 ( [ betagamma ] ) near @xmath56 is infrared stable ( ! ) , in sharp contrast to ( for example ) the impurity quantum critical point of the pseudogap kondo model.@xcite the absence of runaway flow in eq . ( [ betagamma ] ) suggests the absence of a strong - coupling phase with complete screening , even at the bulk critical point with gapless bosonic modes . in a sense , this is not unexpected , as eq . ( [ fluctfield ] ) describes the classical fluctuations of an effective magnetic field coupled to the impurity , being unable to quench the entropy . as a consequence of , no quantum phase transitions occur in the model ( [ fluctfield ] ) . the impurity properties are controlled by a stable intermediate - coupling fixed point , which describes a non - trivially fluctuating fractional - spin state . near this fixed point , the impurity susceptibility and entropy obey@xcite @xmath57 with universal constants @xmath58 , @xmath59 ( which depend on dimensionality @xmath2 and impurity spin size @xmath60 only ) . we note that the problem of an impurity in a collinear magnet can equivalently be formulated using a non - linear sigma model for the bulk@xcite this leads to an expansion in @xmath61 , with similar impurity properties . thus , the fractional - spin state is present for all @xmath62 for @xmath63 the field - theoretic predictions@xcite have been verified by extensive quantum monte carlo simulations.@xcite paranthetically , we note that the fixed point at @xmath56 is unstable w.r.t . breaking of the underlying su(2 ) symmetry : a stable intermediate - coupling fixed point still exists in the xy case , whereas flow to strong coupling occurs in the ising situation.@xcite as above , we intend to couple a magnetic impurity to the local order - parameter field of the bulk , i.e. , by a term @xmath65 , but now the low - energy dynamics of @xmath66 will be represented in terms of spinon fields @xmath67 introduced in sec . [ noncollinearorder ] . it is easy to see that an appropriate choice of a linear combinations of the spinons defined in eq . , namely @xmath68 enables us to write expression in an alternative , but very convenient way : @xmath69 note that this rotation leaves the bulk action invariant . the appropriate model to describe a magnetic impurity in a topological spin liquid with a z@xmath0 gauge structure is thus @xmath70\ ; , \label{equ : bosonickondomodel}\end{aligned}\ ] ] and was proposed recently by us.@xcite the most important feature of model is that the impurity is coupled to a bilinear of the elementary bulk excitations ( i.e. , the spinons ) . this has to be contrasted to the impurity theory with collinear bulk ordering , where the impurity spin couples _ linearly _ to the magnetic ( triplon ) modes . the consequences of this difference will be examined below and turn out to be crucial for the fate of the impurity near the bulk critical point . thus , apart from the strong constraint on the bulk spinons , the model strongly resembles the _ fermionic _ kondo model and constitutes its straightforward extension to a bosonic environment . in the following , we concentrate on the impurity behavior at the bulk critical point , i.e. , for gapless spinons . in the gapped case , weak impurity coupling will be irrelevant , and the impurity will decouple from the bulk in the low - energy limit . however , in analogy to fermionic kondo models with a gapped density of states and broken particle hole symmetry,@xcite a first - order quantum transition will take place upon increasing @xmath71 , beyond which a strong - coupling state is realized . a rg treatment of the action presented in eq . can be performed perturbatively in @xmath71 . in addition , the constraint @xmath72 is relaxed and treated on the mean - field level ( this is equivalent to giving the spinons a mass shift ) . the influence of such bulk interaction terms will be considered in sec . [ bulkint ] . at zero temperature , the renormalized spinon mass vanishes at criticality , and the spinons become soft . this is the regime we focus on , and which is described by the simpler action : @xmath73.\end{aligned}\ ] ] in order to use diagrammatic methods it is convenient to represent the impurity spin in terms of schwinger bosons ( one could equally well choose to work with abrikosov fermions ) . restricting ourselves to an impurity spin @xmath1 , we introduce thus bosonic operators @xmath74 which obey the following constraint @xmath75 this allows for a faithful impurity spin representation according to @xmath76 in terms of these bosonic operators , we can formulate the full impurity action : @xcite @xmath77 where we implement the constraint exactly by taking the limit @xmath78.@xcite in eq . , we have included a potential scattering term @xmath79 on the impurity location @xmath80 . in contrast to the fermionic case , such a term is generated perturbatively in the rg flow . this implies that for finite @xmath81 , @xmath82 can not be a fixed point of the rg this is intimately related to the bosonic nature of the bulk , as bosons can not show particle hole symmetry . the rg treatment starts by determining the tree - level scaling dimensions of the couplings , yielding : @xmath83 = \textrm{dim } [ v_0 ] = d-2 \equiv \epsilon \ ; .\end{aligned}\ ] ] both couplings are marginal in two dimensions ; this allows for a controlled rg procedure , which we shall perform here using the standard field - theoretic scheme . renormalized couplings are introduced as @xmath84 and @xmath85 , where @xmath86 denotes @xmath87 . at two - loop order ( see app . [ appen : bosonrg ] ) we find the following beta functions @xmath88 where @xmath89 for non - interacting bulk bosons . ( this corrects a sign error in the result published in ref . . ) it is important to note that @xmath90 [ @xmath91 does not contain even ( odd ) powers of @xmath81 , respectively . this is the consequence of the invariance of the action w.r.t . the transformation @xmath92 which also implies that ferromagnetic and antiferromagnetic impurity coupling are equivalent . ( note again here that the @xmath67 are not canonical bosons the corresponding operators simply commute . ) the solution of the two - loop flow equations is depicted in fig . [ fig : flow_bose ] . the weak - coupling rg flow has four fixed points : the trivial ( gaussian ) fixed point lm with @xmath93 ; a potential scattering fixed point ps , given by @xmath94 ; and a pair of critical fixed point , located at @xmath95 and denoted qcp . let us quickly discuss the physical implications of the present rg flow ( but note that certain changes will arise from the consideration of bulk interactions , see below ) . for @xmath96 and @xmath97 , the flow indicates two stable phases , namely the local - moment phase ( lm ) , and a strong - coupling phase at large @xmath98 ( the hamiltonian is symmetric under @xmath99 ) and large negative @xmath100 . the two phases are separated by a continuous quantum phase transition , controlled by qcp , with a correlation length exponent given by @xmath101 characterizing the vanishing crossover energy scale near criticality . the pure potential scattering problem , @xmath102 , has two phases as well , which can be easily understood from the exact local spinon green function : @xmath103 values @xmath104 flow to zero ( the denominator in eq . ( [ deno ] ) is regular ) , whereas @xmath105 induces a bound state ( a pole appears in the green function of the bosons ) . the physical properties of lm are that of a decoupled impurity , i.e. , susceptibility and entropy obey @xmath106 , @xmath107 in the low - temperature limit . the strong - coupling regime can not be analyzed using the present rg ; a combination of large-@xmath4 methods and strong - coupling models indicates that true screening can occur ( for both signs of @xmath71 ) , with a finite impurity susceptibility and vanishing entropy.@xcite finally , at the critical point ( qcp ) properties similar to eq . obtain . for @xmath108 , the weak - coupling regime is unstable , and the flow generically is towards strong coupling this can be related to the bare boson density of states , eq . ( [ bare ] ) , diverging as @xmath109 . nevertheless , there is still a qcp , which separates the strong - coupling ( screened ) regime from a pure potential - scattering phase at positive @xmath100 ( controlled by a now stable ps fixed point ) . for @xmath110 no quantum phase transition occurs , and there is logarithmically slow flow to strong coupling ( see below for corrections to this picture coming from bulk interactions ) . overall , the rg flow is strikingly different from an impurity which is coupled to gapless spin-@xmath111 bosons ( sec . [ vojtabura ] ) , where the coupling @xmath45 flows towards a _ stable _ intermediate - coupling fixed point . ; ( b ) @xmath108 . ( @xmath112 for non - interacting spinons ; note that @xmath54 may be non - zero in @xmath63 owing to bulk spinon interactions . ) qcp denotes the boundary quantum critical point and the blue line is the separatrix , separating a strongly coupled phase from either a local moment ( lm ) regime or a potential scattering ( ps ) a zero kondo coupling . the curves have been calculated from for @xmath113 . note that the flow is symmetric w.r.t . , scaledwidth=48.0% ] within the above weak - coupling rg the bulk interactions ( represented by the constraint @xmath115 ) were only treated at the mean - field level . unfortunately , a consistent rg fully including the bulk interactions can not be performed : the appropriate bulk non - linear sigma model is perturbatively accessible near @xmath116 ( while softening the bulk interaction allows an expansion in @xmath117 , see below ) , but expansions around both @xmath116 and @xmath55 are incompatible with the bare scaling dimension of the impurity coupling @xmath71 , which is marginal in @xmath63 . ( this is in contrast to the situation in sec . [ vojtabura ] where _ both _ bulk and boundary couplings were marginal at @xmath55 . ) the treatment of bulk interactions in the present spinon case is , however , physically important , both at weak coupling ( to be discussed below ) and at strong coupling this was discussed in some detail in ref . , but requires further work which is beyond the scope of this paper . let us qualitatively discuss the role of bulk spinon interactions at weak coupling . the scaling dimensions of @xmath81 and @xmath100 are determined by the local bulk spin correlations . in the interacting bulk theory , this decay is given as @xmath118 where @xmath119 is the anomalous dimension . ( note that @xmath120 for non - interacting spinons , whereas @xmath121 in the case of a standard @xmath5 theory controlled by a gaussian fixed point . ) this implies that @xmath122 = \textrm{dim}[v_0 ] = ( d-3+\eta)/2 $ ] , i.e. , the @xmath89 in the rg equations ( [ rgeq ] ) should be replaced by @xmath123 . the qcp in the incommensurate magnet has @xmath124 in @xmath63 ( for details see refs . ) , i.e. , a boundary quantum phase transition as shown in fig . [ fig : flow_bose](a ) is indeed possible for the physical two - dimensional case . we note that this analysis ( i.e. including @xmath119 in the scaling dimensions of @xmath81 and @xmath100 ) is in principle correct at one - loop order ; at higher orders interference processes between bulk and boundary interactions can be expected . no systematic treatment of those is possible in the present formulation . therefore , we propose a more rigorous way to incorporate bulk interactions , which leads us to a bosonic version of the anderson impurity model . our goal is to find a representation of the model which allows to perturbatively control both bulk and boundary interactions at the same time . to this end , we reformulate the impurity part , i.e. , switch from a kondo to an anderson impurity model such a scheme has proven extremely successful for the pseudogap kondo model.@xcite two transformations will convert the original model into a new one , where bulk and boundary interactions have a common critical dimension : ( a ) an inverse schrieffer - wolff transformation ( in analogy to the pseudogap kondo case ) , which transforms the kondo into an anderson model and renders the hybridization between impurity and bulk marginal in @xmath55 ; ( b ) replacing the hard constraint on the spinons by a `` soft '' self - interaction of the type @xmath125 this term is marginal in @xmath55 as well . ( as will become clear , step ( a ) requires the existence of a strong - coupling singlet state with screening in the original model . ) with steps ( a ) and ( b ) we are led to propose the following field theory : @xmath126 \nonumber \\ & & + \int d\tau \left [ \overline{b}_0 \left ( \partial_\tau+\lambda + \epsilon_0 \right)b_0 + \overline{b}_\sigma \left ( \partial_\tau + \lambda \right ) b_\sigma \right ] \nonumber \\ & + & w_0 \int d\tau \left [ \overline{b}_\sigma b_0 z^{\phantom{}}_\sigma+\textrm{h.c.}\right ] . \label{bosoand}\end{aligned}\ ] ] we have left out a bulk mass term @xmath127 , as we are interested in the bulk critical point . the impurity is now a three - level system ( similar to an anderson model with infinite repulsion ) , represented by three auxiliary particles @xmath74 ( @xmath28 ) and @xmath128 note that @xmath128 may be interpreted as the empty impurity state . the impurity transition is tuned by the value of @xmath129 : for @xmath130 the impurity is in a doublet state ( i.e. unscreened ) whereas for @xmath131 the state is a singlet ( these statements apply for @xmath132 , finite @xmath133 will shift the impurity critical point ) . the hilbert space dimension is enforced by the constraint @xmath134 . technically , this is again implemented via a chemical potential @xmath135.@xcite the action incorporates three main ingredients : ( i ) it contains an interaction term @xmath136 in the bulk , which is marginal in @xmath55 ; ( ii ) the hybridization @xmath133 is marginal in @xmath55 as well ; ( iii ) in the `` kondo '' limit , i.e. , for a large positive singlet energy @xmath129 , a schrieffer - wolff transformation allows to integrate out the @xmath128 field , and leads to the previous bosonic kondo interaction ( [ fullmodel ] ) . the model ( [ bosoand ] ) at @xmath55 is thus strongly reminiscent of the pseudogap anderson impurity model ( with infinite impurity coulomb energy ) and linear density of states.@xcite ( another bosonic anderson model was recently introduced by lee and bulla,@xcite which , however , displays different physics , as the model of ref . is spinless , but with finite on - site repulsion . ) we will proceed with deriving the rg flow for the anderson model . the rg flow for both the bulk and impurity interactions , @xmath136 and @xmath133 , can be perturbatively controlled in @xmath137 . this allows a consistent calculation of the critical properties of the fully interacting bosonic impurity model . the details of this calculation are presented in app . [ appen : bosonicandersonrg ] . we introduce renormalized couplings according to @xmath138 and @xmath139 . placing ourselves at both the bulk and boundary critical points ( ! ) , we find the following flow equations to one - loop order : @xmath140 these rg equations describe the non - trivial flow of both bulk and boundary interactions at the critical point in @xmath141 . higher - order contributions ( not shown here ) contain the above - mentioned interference processes and induce a feedback of the bulk interaction @xmath142 into the impurity flow . we refrain from a more detailed analysis at this point : qualitatively , the rg approaches in secs . [ sec : weak ] and [ sec : ande ] yield similar results ; a quantitative comparison is difficult as the two expansions are around a lower critical and upper critical dimension of the impurity problem , respectively . let us , however , emphasize that we have demonstrated how a systematic expansion , analogous to the one in ref . for impurity models in collinear magnets ( i.e. treating bulk and boundary interactions on equal footing ) can be set up in the non - collinear case as well . in this last part , we briefly introduce and analyze the problem of a two - state system@xcite ( described by a magnetic pseudospin impurity ) coupled to a two - component non - relativistic bose gas.@xcite this extends the kind of bosonic impurity models , introduced above in the context of quantum antiferromagnets , to the field of cold atomic gases . technically , the main difference to the situation in the previous section is that the bulk now consists of _ canonical _ bosons . the action assumes the form : @xmath143 \ ; , \label{canmodel}\end{aligned}\ ] ] where we consider canonical bosonic particles , @xmath144 , with two internal degrees of freedom ( @xmath28 here ) and a mass @xmath145 ( i.e. , chemical potential ) . in the field of ultracold trapped gases , usually performed with fixed particle number , @xmath146 in the low - temperature limit . interactions among bulk bosons are neglected ; to lowest order , they lead to a temperature - dependent renormalization of @xmath145 . the impurity couples to the bosonic spin density at site @xmath147 with a kondo - type exchange interaction , but an anderson - like hamiltonian similar to eq . ( [ bosoand ] ) could be considered as well . in contrast to spinon - based models , we realize that there is no invariance under the transformation @xmath148 , i.e. , ferromagnetic and antiferromagnetic impurity coupling are no longer equivalent . however , as in the spinon case , a potential scattering term is generated during the flow even for @xmath149 . in the remainder , we focus on the weak - coupling analysis in the massless case , @xmath150 . ( in the massive case , @xmath151 at @xmath152 , we again expect a first - order transition as function of @xmath71 . ) a tree - level dimensional analysis yields @xmath153=\textrm{dim}[v_0]=d-2\equiv \epsilon , \end{aligned}\ ] ] similar to the model in sec . [ sec : weak ] . we thus introduce dimensionless couplings according to @xmath154 and @xmath155 and find to one - loop order @xmath156 remarkably , this result is exact to _ all _ orders in perturbation theory ( this was previously shown for non - relativistic @xmath5-theory with local interactions@xcite ) . technically , the reason for that lies in the diagrammatic structure of the problem , where only ladder - like diagrams can be different from zero in the limit @xmath152 . higher - order terms are cancelled by the appropriate counterterms . physically , the exactness of eq . is tied to the fact that the expansion is performed around a trivial vacuum state . the rg flow , derived from eq . , is depicted in fig . [ fig : flow_canonicalbose ] . for positive ( antiferromagnetic ) @xmath81 , it is qualitatively similar to the one of the spinon model in sec . [ sec : weak ] . from eq . one finds four fixed points : the local - moment ( lm ) fixed point at @xmath93 , a potential - scattering ( ps ) fixed point at @xmath157 , and two critical fixed points at @xmath158 and @xmath159 . for both qcp , the correlation length exponent is @xmath160 which again is exact to all orders in perturbation theory . the structure of the phase diagram for both positive and negative @xmath54 is thus similar to the spinon case in sec . [ sec : weak ] . however , ferromagnetic and antiferromagnetic @xmath71 are no longer equivalent , suggesting the existence of two distinct strong - coupling phases for @xmath161 and @xmath162 ( where true screening is likely to occur only in the latter case ) . for @xmath110 and @xmath162 , logarithmic flow to strong coupling occurs , with a screening temperature given by @xmath163 as in the conventional kondo problem . ; ( b ) @xmath108 . ( @xmath112 for non - interacting bosons ; again , @xmath54 may be non - zero in @xmath63 due to bulk interactions . ) the curves have been calculated from for @xmath113 . , scaledwidth=47.0% ] thus , the situation of a two - level impurity coupled to a two - component bose gas , with a pseudospin pseudospin coupling , allows for three distinct phases ( given @xmath164 ) : a weakly coupled ( unscreened ) phase and two strong - coupling phases . the present analysis can be easily generalized to an arbitrary number of pseudospin components or larger effective spin sizes ; at weak coupling no qualitative changes are expected . however , the precise nature of the ground state at strong coupling depends on microscopic parameters of the bose gas , such as spin size and the value of the repulsion between bosonic atoms.@xcite more elaborate strong - coupling methods are required to address this issue . last not least , let us point out another remarkable feature of the above model : its rg flow is equivalent to a fermionic kondo problem with full band asymmetry , i.e , with a chemical potential tuned to the band edge . this was considered previously,@xcite and we comment on this point in app . [ appen : fermions ] . in this paper , we have introduced a new class of quantum impurity models , describing the dynamics of a two - level system ( spin ) coupled to an external bosonic spin - carrying environment . two different situations have been studied in the weak - coupling regime , namely a spin @xmath1 magnetic impurity in a class of @xmath64 spin liquids , and a local defect in a two - component bose gas . for massless bulk particles , we have derived the renormalization group flow of the impurity coupling constants . while some differences in the rg flow appear between the above two realizations , the following generic features appear : ( i ) an impurity quantum phase transition is possible which separates a local - moment regime ( decoupled impurity ) from a strongly coupled impurity ; ( ii ) in contrast to a magnetic impurity coupled to a collinear bulk magnet described by a standard @xmath5 theory , there is no stable intermediate - coupling fixed point , but instead run - away flow to strong coupling . ( for gapped bulk particles , a first - order transition will generically appear , in analogy to gapped fermionic kondo models . ) at strong coupling , the nature of the ground state depends on microscopic details ( and was not investigated further here ) , but quenching of the impurity entropy may obtain.@xcite in summary , our results underline that the dynamics of impurity degrees of freedom depends crucially on the nature of the elementary excitations of the surrounding environment . hence , impurity properties may be used as probes for exotic bulk excitations . on the experimental side , cs@xmath0cucl@xmath16 has been proposed to be a realization of a spin-1/2 frustrated antiferromagnet on an anisotropic triangular lattice , which possibly features spinon excitations.@xcite studying the physics of dilute magnetic impurities in this material , e.g. the temperature dependence of nmr knight shifts or of the impurity contribution to the susceptibility , will clearly help to resolve the nature of the bulk magnetic state . we thank s. sachdev , t. senthil and g. zarnd for valuable discussions , and k. damle for a collaboration at the early stage of this work . furthermore , we would like to acknowledge the creators of jaxodraw@xcite , which simplified the drawing of the feynman diagrams significantly . this research was supported by the dfg through the center for functional nanostructures ( karlsruhe ) and sfb 608 ( kln ) , as well as through the virtual quantum phase transitions institute in karlsruhe . in this appendix we derive the rg equations for the bosonic kondo model with bulk spinons ( the derivation is kept in a form which is also appropriate to treat the case of canonical bosons as well as electrons ; the only occurring modifications manifest themselves in the values of the corresponding feynman diagrams ) . the rg equations of bosons are very different form those of electrons , which is the reason why we present a rather detailed analysis . the crucial difference resides in the fact that during the renormalization process a potential scattering term is generated . in order to properly deal with this complication we start with a generalized version of the model proposed in eq . which allows for fully spin - anisotropic couplings this enables us to deal with the kondo interaction and the potential scattering term at a time without worrying about newly generated terms . the action we analyze has the following very general form @xmath165 the tree - level scaling dimensions of the general coupling constants are @xmath166=\textrm{dim}[\beta_0]=\textrm{dim}[\gamma_0]=d-2=\epsilon \ ; .\end{aligned}\ ] ] based on this analysis we introduce dimensionless couplings according to @xmath167 where @xmath48 denotes an arbitrary but fixed energy scale ( the renormalization scale ) and @xmath168 , where @xmath169 with @xmath170 being the euler function . we furthermore state that during the course of renormalization the schwinger boson field parametrizing the spin is renormalized according to @xmath171 we do not have to introduce renormalized fields for the bulk degrees of freedom since the influence of a single impurity on the bulk is an effect of the order @xmath172 , which does not renormalize bulk degrees of freedom . the prefactor @xmath173 can trivially be taken care of by rescaling the fields in eq . according to @xmath174 which has the only advantage of getting rid of bothersome factors in the course of the calculation . the generalized renormalized action ( with appropriate counterterms ) reads : @xmath175 this is the appropriate starting point to iteratively construct the counterterms by demanding the action to be finite at a certain arbitrary but fixed renormalization point . the graphical representation of the vertices and the counterterms is shown in fig . [ vertices ] . the vertex @xmath176 is denoted as a small black dot whereas the associated counterterm @xmath177 is represented as a big black dot in a box . the interaction vertex @xmath178 and its counterterm @xmath179 are represented as a small grey dot and a big grey dot with a cross in the interior , whereas @xmath45 and the associated counterterm @xmath180 are graphically represented as small grey box and a big grey box with a cross inside , respectively . is denoted as a black circle , its counterterm is represented by a big black circle in a box ; the small grey dot and the big grey dot with a cross represent @xmath178 and its counterterm ; the small ( big ) grey box ( grey box with a cross ) represents @xmath45 ( the counterterm of @xmath45 ) . , scaledwidth=48.0% ] in the diagrammatic expansion this enforces us to consider 6 interaction vertices for a consistent treatment . the next step is to specify the renormalization conditions which will give the theory a finite limit once the cutoff is sent to infinity according to @xmath181 where the subscript denotes the appropriate renormalization point , which we do not specify here . the first line fixes the impurity propagator , whereas the second line fixes the vertex functions . we once again stress the fact that the bulk properties will not receive singular corrections in any order of perturbation theory . to lowest loop order we can concentrate on the renormalization of the vertex functions since there is no propagator renormalization at one loop level ( they will play a role at two loop level ) . to lowest loop order there are no propagator renormalizations . we choose a graphical representation of the renormalization conditions fixed in eq . . the renormalization conditions for the vertex functions are shown in fig . [ fig : gammaalpha1loop ] , fig . [ fig : gammabeta1loop ] , and fig . [ fig : gammagamma1loop ] . structurally there are only two types of diagrams occurring , whose most divergent part in an expansion in @xmath182 yields @xmath183 by dressing the integral values with the appropriate interaction vertices we can determine the @xmath184 factors to lowest order as @xmath185 ( big black dot in box ) . , scaledwidth=33.0% ] ( big grey dot with a cross ) . , scaledwidth=33.0% ] ( big grey box with a cross ) . , scaledwidth=33.0% ] in order to derive the @xmath184 factors to two - loop order we only calculate the numerical values of the topologically distinct diagrams shown in fig . [ fig : gammaalpha ] . the actual vertex corrections are obtained by properly dressing the numerical value of the integrals with the corresponding interaction vertices . all possible diagrams contributing to the two - loop renormalization factors are shown in figs . [ fig : gammaalpha ] and [ fig : gammabeta ] . the diagrams contributing to @xmath186 are not shown explicitly , since they can be obtained from the diagrams contributing to @xmath187 by replacing grey dots by grey boxes . we start with the trivial diagrams @xmath188 and @xmath189 which are just the squares of the one - loop diagrams @xmath190 the diagrams @xmath191 are given by @xmath192 which to leading order yields @xmath193 the remaining diagram @xmath194 evaluates to @xmath195 we will re - encounter the same integral within the calculation of the field renormalization factor . in order to arrive at the final two loop @xmath184-factors we dress the integral values with the appropriate interaction vertices . after some calculation we find @xmath196 the next step consists of the propagator renormalization , whose recursion relation is shown in fig . [ fig : bosonself ] . we calculate the diagram shown in fig . [ fig : bosonself ] ( a ) . the dependence on the external frequency can be extracted by calculating @xmath197 ( crossed dot ) . , scaledwidth=48.0% ] if we dress this integral with the appropriate vertices and apply the renormalization condition shown in fig . [ fig : bosonself ] ( note that @xmath198 ) we find @xmath199 in order to derive the rg beta functions we use the fact that the bare couplings have no knowledge about the renormalization condition , which can be expressed as @xmath200 this leads to a set of coupled equations @xmath201 where the introduced @xmath178 functions correspond to @xmath202 etc . these coupled equations can be solved iteratively , leading to @xmath203 in order to arrive at the actual bosonic kondo model ( eq . ) with potential scattering we have to make the following identifications @xmath204 the flow equations of the physical couplings @xmath205 are related to the intermediate quantities @xmath206 according to @xmath207 \ ; , \nonumber \\ \left [ \beta(\beta)+\beta(\gamma ) \right ] \ ; , \end{aligned}\ ] ] with the interaction parameters chosen as shown in eq . . this leads to the following flow equations @xmath208 or in the spin - isotropic situation @xmath209 which are the rg equations quoted in the main text . the rg equations for the model of canonical bosons , eq . in sec . [ canonical ] , are obtained from the same diagrammatic expansion ( albeit with different numerical values for the diagrams ) . the observation that the rg equations are exact to _ all _ orders comes can be easily understood by noting that the counterterms cancel all higher - order contributions because the only non - vanishing diagrams are ladder - like ( in electronic language one would call these diagrams the cooper diagrams ) , see ref . . within this section we derive the rg equations for the bosonic version of the infinite-@xmath210 anderson model of sec . [ sec : boseanderson ] . after a fourier transformation for the quadratic bulk part , the action of the model reads @xmath211 z_\sigma(\nu_n , k)\nonumber \\ & + & u_0 \int d^d x d\tau z^_\sigma z_\sigma z^*_{\sigma'}z_{\sigma ' } \nonumber \\ & + & \int d\tau \overline{b}_0(\partial_\tau+\lambda)b_0 + \int d\tau \overline{b}_\sigma ( \partial_\tau+\lambda)b_\sigma \nonumber \\ & + & w_0 \int d\tau \left [ \overline{b}_\sigma b_0 z_\sigma ( \tau ) + \textrm{h.c . } \right].\end{aligned}\ ] ] the flow equations for the bulk interaction @xmath136 and the hybridization @xmath133 will not mix to one - loop order . in the following we will ignore the renormalization of the mass difference between the different impurity levels [ @xmath129 in ] , i.e. , we work at criticality . we introduce renormalized couplings according to @xmath212 where @xmath137 . it is interesting to note that technically we have a version of the rg of a model which is at its upper critical dimension ( both for the bulk and the impurity ) , whereas the rg for the bosonic kondo model is reminiscent of an rg at the lower critical dimension . the vertices of the theory are shown in fig . [ fig : vertexdef ] . , the dashed line is @xmath74 and the full line stands for @xmath31 ( incoming arrow ) or @xmath213 ( outgoing arrow ) . ( b ) and ( c ) denote self - energy corrections to the local levels @xmath74 ( b ) and @xmath128 ( c ) . , scaledwidth=40.0% ] the calculation for the renormalization of the hybridization @xmath214 completely parallels the analysis in the pseudogap infinite-@xmath210 anderson model , which was presented in ref . , and therefore without showing further calculational details leads to the same flow equation , namely @xmath215 in addition , we have to calculate the perturbative corrections to the bulk interaction vertex , which is identical to the usual o(4)-@xmath5 theory . in our convention the result is : @xmath216 the hidden o(4 ) symmetry of the theory becomes apparent upon writing the bosonic spinons as a 4-vector @xmath217 . remarkably , the two - level system coupled to the two - component bosonic gas studied in sec . [ canonical ] turns out to follow a weak - coupling rg flow indentical to that of a fermionic kondo problem , where the chemical potential is tuned to the ( e.g. lower ) band edge ( i.e. with full particle - hole asymmetry ) . the equivalence can best be understood in terms of the local density of states the only quantity which determines the impurity behavior which in both cases ( canonical bosons and fermions ) is of the form @xmath218 . the strict particle - hole asymmetry of the electrons implies that there are only particle excitations . introducing dimensionless couplings according to standard practice@xcite we can derive the following flow equations for the fermionic problem : @xmath219 these equations are contained in the earlier analysis of ref . , and are identical to those obtained for the problem of canonical bosons in sec . [ canonical ] . again , the equations are exact to all orders in perturbation theory ( ! ) . for a general fermionic metallic ( @xmath220 ) kondo problem with band asymmetry ( i.e. the chemical potential not being pinned to the center of the band ) one has to perform a two - step rg procedure : initially , one has to follow the fully asymmetric scaling , eq .. once the remainder of the band is symmetric , the usual poor man s scaling equations take over here the potential scattering term is a marginal operator . we note that the fully asymmetric fermionic kondo problem can be relevant in the context of dmft for hubbard models : the transition from the undoped to the doped mott insulator upon variation of the chemical potential happens precisely when the chemical potential reaches the gap edge ( provided the transition is continuous ) . note that our notations for conjugate objects , @xmath221 and @xmath222 , aim at distinguishing canonical bosons @xmath223 from commuting numbers @xmath67 . of course , in the language of coherent - state path integrals the difference is in the dynamic term .	we investigate quantum impurity problems , where a local magnetic moment is coupled to the spin density of a bosonic environment , leading to bosonic versions of the standard kondo and anderson impurity models . in a physical situation , these bosonic environments can correspond either to deconfined spinons in certain classes of z@xmath0 frustrated antiferromagnets , or to particles in a multicomponent bose gase ( in which case the spin degree of freedom is attributed to hyperfine levels ) . using renormalization group techniques , we establish that our impurity models , which feature an exchange interaction analogous to kondo impurities in fermi liquids , allow the flow towards a stable strong - coupling state . since the low - energy bosons live around a single point in momentum space , and there is no fermi surface , an impurity quantum phase transition occurs at intermediate coupling , separating screened and unscreened phases . this behavior is qualitatively different from previously studied spin - isotropic variants of the spin - boson model , which display stable intermediate - coupling fixed points and no screening .
universal scaling laws for the dimensional properties of flexible polymer chain in a good solvent , such as the end - to - end distance @xmath4 and the gyration radius @xmath5 , are well explained and understood since the groundbreaking studies by de gennes and des cloizeaux @xcite . it was found that both properties scale as @xmath6 for large enough numbers of monomers @xmath1 , where the exponent @xmath3 is universal and depends on the dimension of space @xmath7 only . besides that , the probability distribution @xmath8 is also examined in both asymoptotic regimes of small and large values of @xmath4 based on the number of chain conformations @xcite . similar arguments allowed lhuillier to suggest an heuristic form for the probability distribution @xmath9 as well @xcite . as far as the scaling laws are also valid for a number of other characteristics of the polymer chain , the universality and scaling properties are often treated as identical concepts . this , however , is a misconception as far as there exist characteristics that are universal but do not obey scaling laws . an example is given by the shape characteristics of polymer chains that which is in the focus of this paper . these are important in a number of applications , to mention here various catalytic activities @xcite and gel chromatography @xcite . the fact that the shape of a polymer coil in a good solvent is not spherical is known since a classical work by kuhn @xcite . however , the understanding that certain shape properties of polymer chains are universal and depend , like scaling parameters , solely on @xmath7 , is brought _ via _ the use of the renormalization group method ( see , e.g. @xcite ) . these findings were also supported by a number of numerical simulations performed on lattice models of the self avoiding walk ( saw ) @xcite . one should note that the mean values for most shape characteristics are different but nonetheless close to their random walk ( rw ) counterparts . some important implication here can be seen in the fact that the probability distributions for most of these characteristics are very broad and asymmetric yielding an ambiguity in the definition of their mean values . the shape of these distributions remains to be explained in a manner similar to the one given for their counterparts @xmath8 @xcite and @xmath9@xcite . most of the simulation studies mentioned above use the monte carlo algorithm applied to the saw lattice model . this approach achieves very good configuration statistics by means of relatively low computation cost . atomistic off - lattice models , on the other hand , allow one to include the effects of chain stiffness , chain composition , the role of solvent , etc . in a chemical way , but at an increased simulation time cost . a good compromise here is the use of coarse - grained approaches that combine the best of two worlds : chemical versatility and computational efficiency . one of such approaches is the dissipative particle dynamics ( dpd ) method @xcite , that has already been used by a number of authors @xcite including two of the current authors @xcite , to examine the scaling properties of a polymer chain in a solvent of variable quality . the benefit of this approach is that it provides the means to study a number of important problems related to microphase separation of amphi- and polyphilic molecules , self - assembly , adsorption , etc . ( see , e.g. @xcite ) . macromolecular shape plays an important role in all of the above mentioned problems . therefore , the validation of this approach with respect to its reliability in predicting the correct macromolecular shape is of high practical interest . this is exactly at the heart of the current study , were we apply dpd simulations to the simplest case a single chain in a good solvent . we examine the asphericity , prolateness and a number of the other shape characteristics , as well as their probability distributions . besides the mean values for each of these properties , we also evaluate their most probable values by means of fitting their probability distributions to respective analytic expressions . the outline of the study is as follows . the simulation approach and the properties of interest are described in sec . [ ii ] , scaling properties and a probability distributions analysis are covered in sec . [ iii ] , probability distributions for shape characteristics are discussed in sec . [ iv ] and conclusions are provided in sec . our study is based on mesoscopic dpd simulations , which has two main advantages : ( i ) neglecting less important properties of a system on small length- and time - scales , and ( ii ) preserving hydrodynamic limits @xcite . a single polymer chain and an explicit solvent are contained within a cubic simulation box with a linear size of at least @xmath10 . here , @xmath11 is the estimate for the gyration radius of a chain of @xmath1 monomers in a good solvent : @xmath12^{0.59}$ ] ( @xmath13 is an estimate for an average bond length as reported in @xcite ) . the monomers are soft beads of equal size , and each one represents either a fragment of a real polymer chain or a few molecules of a solvent . we restrict our study to the case of an athermal solvent , where all types of pairwise bead - bead interactions : polymer - polymer , solvent - solvent and polymer - solvent are identical . the monomer coordinates @xmath14 are defined in continuous space ( in contrary to most studies using the monte carlo method for similar studies , which , typically , are performed on a lattice ) . this allows to describe the phase space of conformations by chains of shorter length . on the other hand , the soft nature of dpd interaction prevents the system from being caught in a metastable state ( what is often observed in the case of the molecular dynamics when atomistic potentials are applied ) . we follow the dpd approach as described in ref . the length is represented in units of the diameter of the soft bead , and the energy scale is assumed to be @xmath15 , where @xmath16 is the boltzmann constant , @xmath17 is the temperature and time is expressed in @xmath18 . the monomers are connected _ via _ harmonic springs , which results in a force @xmath19 where @xmath20 and @xmath21 is the spring constant . the non - bonded forces contain three contributions @xmath22 where @xmath23 is the conservative force , resulting from the repulsion between @xmath24-th and @xmath25-th soft beads , @xmath26 is the dissipative force , that occurs due to the friction between soft beads , and random force @xmath27 that works in pair with a dissipative force to thermostat the system . the expressions for all these three contributions are given below @xcite @xmath28 @xmath29 @xmath30 where @xmath31 , @xmath32 , @xmath33 is the velocity of @xmath24th bead , @xmath34 is the amplitude for the conservative repulsive force . the dissipative force has an amplitude @xmath35 and decays with the distance according to the weight function @xmath36 . the amplitude for the random force is @xmath37 and the respective weight function is @xmath38 . @xmath39 is the gaussian random variable and @xmath40 is the time - step of the simulations . as was shown by espaol and warren @xcite , to satisfy the detailed balance requirement , the amplitudes and weight functions for the dissipative and random forces should be interrelated : @xmath41 and @xmath42 ^ 2 $ ] . the system is soft repulsive and is kept together by an external pressure to provide the required density , which corresponds to a liquid state . in these simulations the following numeric values are used : @xmath43 for all pairs of interacting beads , @xmath44 , @xmath45 and the time - step @xmath46 . the duration of all runs ( each performed for a different chain length @xmath47 ) was fixed at @xmath48 dpd steps . here and thereafter we consider the case of space dimension @xmath49 only . all shape characteristics of a chain are derived from the components of the instantaneous gyration tensor @xmath50 defined as in @xcite : @xmath51 here , @xmath1 is the number of monomers of a chain , @xmath52 denotes the set of the cartesian coordinates of @xmath53th monomer center : @xmath54 , and @xmath55 are the coordinates of the center of mass for the chain . its eigenvectors define the axes of a local frame of a chain and the mass distribution of the latter along each axis is given by the respective eigenvalue @xmath56 , @xmath57 , respectively . the trace of @xmath50 is an invariant with respect to rotations and is equal to an instantaneous squared gyration radius of the chain @xmath58 here , the average over three eigenvalues , @xmath59 , is introduced to simplify the following expressions . the instantaneous asphericity @xmath60 ( sometimes also referred to as `` the relative shape anisotropy '' ) . the prolateness @xmath61 and size ratio @xmath62 are defined as @xcite @xmath63 where @xmath4 is the magnitude of the end - to - end vector @xmath64 . a spherical shape is characterised by @xmath65 , whereas for a non - spherical one it is : @xmath66 and @xmath67 ( for prolate shape ) and @xmath68 ( for oblate shapes ) . to define the remaining shape characteristics we follow ref . @xcite . to this end we introduce the following triplet of vectors : @xmath69 , @xmath70 and @xmath71 . here @xmath72 , @xmath70 is the component of the vector @xmath73 perpendicular to @xmath69 , whereas @xmath71 is the component of the vector @xmath74 perpendicular to both @xmath69 and @xmath70 . for more details , fig . 1 in ref . @xcite . on a technical note , for the case of a non - integer @xmath75 ( or @xmath76 ) , the vector @xmath77 ( or @xmath78 ) was chosen _ via _ interpolation between the vectors @xmath79 and @xmath80 for the adjacent monomers , where @xmath81 [ or @xmath82 . the magnitudes of @xmath69 , @xmath70 and @xmath71 are denoted as @xmath83 , @xmath84 and @xmath85 , respectively . then , the ratios @xmath86 are evaluated at each time instance . in real experiments one observes shape characteristics of polymer chains averaged over a sample and over time trajectory . the same can be done in the course of the dpd simulation study . we will denote this hereafter as @xmath87 . let us note that all expressions [ eqs . ( [ asg_def ] ) ( [ r12_r13_def ] ) ] contain these ratios . therefore , the averaging can be performed in two ways : as a ratio of the averages or , alternatively , as the average of the ratios , both ways are used in the literature @xcite . the former definition leads to the first set of the averages : @xmath88 @xmath89 whereas the latter definition yields a second set of averages : @xmath90 @xmath91 ( a ) log - log plot for the average eigenvalues @xmath92 of the gyration tensor , the squared gyration radius @xmath93 and the squared end - to - end distance @xmath94 . ( b ) the same for the average values of asphericity and prolateness nominators and denominators as defined in eq . ( [ as_nom_den ] ) . @xmath1 is the number of monomers per chain . note , that different angles , observed for the two sets of curves in this figure are due to the fact that scaling of @xmath95 and @xmath96 is governed by the exponent 4@xmath3 , whereas that of @xmath97 and @xmath98 by @xmath99 , see eq . ( [ nom_den_scl]).,title="fig:",width=302 ] ( a ) log - log plot for the average eigenvalues @xmath92 of the gyration tensor , the squared gyration radius @xmath93 and the squared end - to - end distance @xmath94 . ( b ) the same for the average values of asphericity and prolateness nominators and denominators as defined in eq . ( [ as_nom_den ] ) . @xmath1 is the number of monomers per chain . note , that different angles , observed for the two sets of curves in this figure are due to the fact that scaling of @xmath95 and @xmath96 is governed by the exponent 4@xmath3 , whereas that of @xmath97 and @xmath98 by @xmath99 , see eq . ( [ nom_den_scl]).,title="fig:",width=302 ] the scaling laws ( [ re_rg_scl ] ) for the polymer chain in a good solvent modelled by means of the dpd method have previously been discussed in detail @xcite . therefore , we will recall these rather briefly here . in particular , as indicated in fig . [ scaling_l123rgsq ] ( a ) , at @xmath2 both @xmath100 and @xmath94 obey the expected scaling laws ( [ re_rg_scl ] ) reasonably well yielding an estimate for the flory exponent falling into the interval of @xmath101 . this is centered around the best known estimate , @xmath102 , obtained by means of the renormalisation group approach @xcite . as far as the asphericity @xmath60 and the prolateness @xmath61 of a polymer chain are defined _ via _ combinations of eigenvalues @xmath103 of the gyration tensor @xmath50 [ see , eq . [ asg_def ] ] , the scaling properties of @xmath103 are also of much interest . such an analysis for the case of the saw was performed first in ref . @xcite using lattice mc simulations . it has been demonstrated that the eigenvalues @xmath103 obey the same scaling laws as the traditional global observables , i.e. @xmath4 and @xmath5 @xmath104 and that the corrections to scaling for the eigenvalues are different but have the same sign as their counterparts for @xmath5 . in our dpd simulations we also find that @xmath103 obey the same scaling laws as @xmath5 , see , respective legends in fig . [ scaling_l123rgsq ] ( a ) . the respective exponents @xmath3 for each eigenvalue , obtained _ via _ linear fit of the data for @xmath2 , are indicated in the same figure and are found to be within the interval @xmath105 . this spread of values must be attributed to different magnitudes of the correction to scaling terms @xcite in each case . these terms are not accounted for in the current study , as far as accurate values for @xmath3 are not the primary goal of this study . therefore , within the accuracy of our simulations , one may state that the same scaling properties are obeyed by all eigenvalues @xmath103 of the gyration tensor @xmath50 . this can be interpreted as an isotropicity of the self - similarity properties of a polymer chain . due to this statement , and given the definitions for @xmath60 and @xmath61 ( [ asg_def ] ) , these shape characteristics are expected to be independent of @xmath1 . to have an additional numerical confirmation for this to hold , we consider the respective nominators and denominators @xmath106 denoted here as `` nom '' and `` den '' . taking into account eq . ( [ lambda_scl ] ) , one expects the following scaling behaviour @xmath107 the data shown in fig . [ scaling_l123rgsq ] ( b ) indicates that this scaling behaviour holds well for @xmath2 for all four properties defined in eq . ( [ as_nom_den ] ) yielding the values of the flory exponent @xmath3 ( see the figure ) consistent with their counterparts for @xmath5 and @xmath103 shown in fig . [ scaling_l123rgsq ] ( a ) . ( a ) dependence of the average values @xmath108 and @xmath109 on @xmath1 . ( b ) the same for @xmath110 and @xmath111 . dashed lines indicate the interval of the averaging for evaluating the final mean value.,title="fig:",width=302 ] ( a ) dependence of the average values @xmath108 and @xmath109 on @xmath1 . ( b ) the same for @xmath110 and @xmath111 . dashed lines indicate the interval of the averaging for evaluating the final mean value.,title="fig:",width=302 ] as a consequence of this , the average values @xmath108 , @xmath109 , @xmath110 and @xmath111 all demonstrate a weak dependence on @xmath1 for @xmath2 , as shown in fig . [ scaling_as ] . the final mean values for these characteristics are , therefore , obtained by averaging the data within an interval @xmath112 , as indicated by dashed lines in both frames of this figure . ( a ) dependence of the average values @xmath113 , @xmath114 and @xmath115 on @xmath1 . ( b ) the same for @xmath116 , @xmath117 and @xmath118 . dashed lines indicate the interval of the averaging for evaluating the final mean value ( where applies).,title="fig:",width=302 ] ( a ) dependence of the average values @xmath113 , @xmath114 and @xmath115 on @xmath1 . ( b ) the same for @xmath116 , @xmath117 and @xmath118 . dashed lines indicate the interval of the averaging for evaluating the final mean value ( where applies).,title="fig:",width=302 ] the situation is markedly different for other shape characteristics , where essential @xmath1 dependence is observed , see fig . [ scaling_gr12r13 ] . the values for @xmath113 and @xmath119 demonstrate steady monotonic growth within the whole interval @xmath120 considered in this study . @xmath121 and @xmath118 stabilize more or less only at @xmath122 , where the estimate for an average is made . the interval for @xmath123 and @xmath117 is a slightly broader : @xmath124 . the mean values for the shape properties defined in eqs . ( [ asg_hat])-([r12_r13_av ] ) and evaluated in our study as discussed above , are collected in column dpd of table . [ tab_aver ] . here , we also list the results obtained for saws by means of other methods , namely , the direct renormalisation approach [ rg ( dr ) ] and monte carlo studies ( mc ) , as well as the respective values for the case of random walks ( rw ) . let us consider @xmath108 and @xmath109 first . one should note that the previous results reported for these characteristics for the case of the saw , are found to be very close to their counterparts for the rw case . therefore , a high accuracy is needed to distinguish between both sets unambiguously . in our case of an off - lattice dpd simulations with explicit solvent , this would be very computationally demanding . nevertheless , within the accuracy limitation of this study , the values found for both @xmath108 and @xmath109 agree reasonably well with the best respective estimates made for the case of saws . for the case of @xmath110 and @xmath111 , the difference between the respective values for saw and rw are more essential . we find reasonably good agreement between our results and other data for saw in as demonstrated in fig . [ scaling_gr12r13 ] . .[tab_aver]the results for mean values of asphericity @xmath108 , @xmath110 , prolateness @xmath109 , @xmath125 and size ratios @xmath113 , @xmath116 , @xmath114 , @xmath117 , @xmath115 , @xmath126 defined according to eqs . ( [ asg_hat])-([r12_r13_av ] ) and their respective most probable values @xmath127 , @xmath128 , @xmath129 and @xmath130 . the abbreviations stand for : random walk ( rw ) , self avoiding walk ( saw ) , lattice monte carlo ( mc ) , direct renormalization ( dr ) , dissipative particle dynamics ( dpd ) ( this work ) . shorthands for citations : @xmath131 @xcite , @xmath132 @xcite , @xmath133 @xcite , @xmath134 @xcite , @xmath135 @xcite , @xmath136 @xcite . data obtained _ via _ fits by eqs . ( [ gauss ] ) , ( [ lhuillier ] ) and ( [ lhuillier2 ] ) are denoted as [ g ] , [ l ] and [ l@xmath137 , respectively . [ cols="^,<,<,<,<,<",options="header " , ] it has been observed before @xcite that the probability distributions of the shape characteristics of a polymer chain are broad and skewed . this is true for both cases of the rw and saw and `` implies that any description of the shapes of random walks which is based only on mean values of related magnitudes is incomplete '' @xcite . one of the ways to extend the analysis of the shape characteristics is to complement the mean values by additional characteristic values obtained from the respective probability distributions . probability distribution @xmath138 for reduced end - to - end distance @xmath139 . ( a ) log - log plot for fitting the asymptotics at @xmath140 . ( b ) double log plot for fitting the asymptotics at @xmath141 built for various trial constant @xmath142 , see eq . ( [ re_dist_asympt ] ) and the text for more details.,title="fig:",width=207 ] probability distribution @xmath138 for reduced end - to - end distance @xmath139 . ( a ) log - log plot for fitting the asymptotics at @xmath140 . ( b ) double log plot for fitting the asymptotics at @xmath141 built for various trial constant @xmath142 , see eq . ( [ re_dist_asympt ] ) and the text for more details.,title="fig:",width=207 ] before we proceed to the probability distributions for the shape characteristics , let us check known asymptotics for the distributions of the end - to - end distance and the gyration radius . to this end we introduce respective reduced properties @xmath143 as has been shown in @xcite , the probability distribution @xmath138 has the following asymptotics : @xmath144 $ , & ~~at~~$r_e^\gg 0$. \end{tabular } \right.\ ] ] where @xmath60 and @xmath142 are ( non - universal ) constants , and @xmath145 . we built a single cumulative histogram for the probability distribution @xmath138 based on simulation data within a scaling regime , i.e. @xmath112 . to examine its asymptotics at @xmath140 , it was plotted in a log - log scale , see fig . [ re_dist_lr ] ( a ) . the @xmath146 power law is displayed _ via _ dashed line and is marked as @xmath140 label . it holds approximately but suffers from somewhat insufficient accuracy near the tail of the distribution ( where the statistics is the poorest ) . however , one also observes a large region of @xmath139 values where another power law holds well , namely @xmath147 with @xmath148 . to check for the asymptotics of @xmath138 at @xmath141 , the logarithm of the histogram @xmath138 is plotted now scaled by the factor @xmath142 . as this factor is not known _ a priori _ , we plotted a family of curves using various values for @xmath142 , see fig . [ re_dist_lr ] ( b ) . all curves converge to the exponential asymptotics in eq . ( [ re_dist_asympt ] ) , which is shown in the figure _ via _ a dashed black line . one may conclude therefore that the known asymptotics for the probability distribution @xmath138 are adequately reproduced in our simulations . probability distribution @xmath138 and the results of fits according to eq . ( [ gauss ] ) ( marked as [ g ] ) and eq . ( [ lhuillier ] ) ( marked as [ l ] ) . the parameters of fits are shown in a figure.,title="fig:",width=264 ] + one should note that the asymptotic regimes ( [ re_dist_asympt ] ) are observed rather at extreme values of @xmath139 . therefore , it is of interest whether or not the probability distribution @xmath138 can be fitted _ via _ appropriate analytic expressions reasonably well within a whole interval of @xmath139 values . the form for @xmath138 , as seen in fig . [ re_dist ] , appears to be weakly asymmetric . therefore , one of the obvious analytic expressions to apply is a generalised gaussian one @xmath149,\;\;\;\bar{x}=x_0.\ ] ] here and thereafter , the most probable value is denoted as @xmath150 . this fit is marked as [ g ] in fig . [ re_dist ] and yields the exponent @xmath151 and the most probable value of @xmath152 . one should remark that this fit closely resembles a standard gaussian distribution ( achieved for parameters values @xmath153 and @xmath154 ) . another choice can be terms of a lhuillier - like form ( for more details , see below ) @xmath155,\;\;\ ; \bar{x}=x'\big(\frac{\epsilon_1}{\epsilon_2}\big)^{\frac{1}{\epsilon_1+\epsilon_2}}.\ ] ] this fit is marked as [ l ] in fig . [ re_dist ] yielding @xmath156 , @xmath157 and @xmath158 . one may conclude that the values for @xmath159 obtained in both fits are very close to the mean value for the scaled end - to - end distance which is equal to one . most studies on probability distributions of shape characteristics of polymer chains are undertaken for the case of a gaussian chain , which mimics the rw @xcite . in this case the analytic evaluation is possible , as well as the @xmath160 expansion , all leading to certain analytic expressions for the probability distributions @xmath161 and , in some cases , for @xmath162 @xcite . on the other hand , in most studies that address such probability distributions for the saw case @xcite , the authors mainly concentrate on broad shapes of such distributions and compare these with their counterparts for the rw . while an analytic solution for the saw is not available , there are several indirect options suggesting certain analytic expressions for the probability distributions of interest . one path is suggested by sciutto @xcite , where the chi - squared form of the distributions @xmath161 are obtained analytically for the rw is transferred to the saw case with the different set of parameter values . while the result is very satisfactory for the eigenvalues ratios @xmath163 , the probability distributions @xmath162 and @xmath164 are not fitted accurately . probability distribution @xmath165 and the results of a fit according to eq . ( [ lhuillier ] ) with the exponents @xmath166 and @xmath167 . most probable value and flory exponent @xmath3 as the result of a fit are shown.,title="fig:",width=264 ] + another approach has been suggested by lhuillier @xcite , where the empirical form ( [ lhuillier ] ) has been deduced for the probability distribution @xmath165 . in particular , the authors consider statistical weights for extreme cases of collapsed ( @xmath168 ) and highly stretched ( @xmath169 ) conformations resulting in the form ( [ lhuillier ] ) for the probability distribution @xmath165 . both exponents @xmath166 and @xmath167 are found to depend solely on the flory exponent @xmath3 ( for the three - dimensional case ) . this , therefore , opens up a possibility to have another independent estimate for the exponent @xmath3 performing a fit of @xmath165 obtained in simulations to the form ( [ lhuillier ] ) taking into account the expressions for @xmath170 and @xmath171 as functions of @xmath3 . applying this analytic expression ( [ lhuillier ] ) to our simulation data we find it to work extremely well , see fig . [ rg_dist ] , yielding the estimate @xmath172 for the flory exponent . it is consistent with previous estimates provided in figs.[scaling_l123rgsq ] and [ scaling_as ] . the most probable value is @xmath173 which deviates for about 10% from the mean value equal to one , as the result of an essential asymmetry of the distribution . due to this asymmetry , the fit used in eq.([gauss ] ) makes essentially no sense . probability distributions for the eigenvalues @xmath103 and squared radius of gyration @xmath5 . ( a ) distributions for unscaled variables . ( b ) distributions for scaled variables , @xmath174.,title="fig:",width=188 ] probability distributions for the eigenvalues @xmath103 and squared radius of gyration @xmath5 . ( a ) distributions for unscaled variables . ( b ) distributions for scaled variables , @xmath174.,title="fig:",width=188 ] in a further approach , one may consider the probability distributions for the eigenvalues @xmath175 of the gyration tensor @xmath50 shown in fig . [ l123_dist ] ( a ) . for the sake of comparison , we also display the probability distribution for their sum , which is equal to @xmath176 . one should note an increase width of the distributions corresponding to an increase of respective eigenvalues @xmath177 . this fact has been already discussed in a number of studies @xcite . another important point whether or not the probability distributions for their scaled counterparts @xmath178 can be mapped onto a single master curve has not discussed so far , except for the case of the rw @xcite . there the distributions @xmath179 are found not to coincide . our respective simulation data are displayed in fig . [ l123_dist ] ( b ) , from which it appears that : ( i ) the probability distributions @xmath180 , @xmath181 and @xmath182 $ ] overlap very closely with little or none deviation , and ( ii ) while @xmath183 does not . currently , we do not have single physical explanation for point ( i ) but point ( ii ) can be explained easily . let us go back to fig . [ l123_dist ] ( a ) . in a limit of a highly stretched chain one has @xmath184 and @xmath185 . hence , at large enough @xmath176 and @xmath186 , their respective distributions should overlap . on the other hand , in a most probable , coil state , the contribution to @xmath176 from @xmath187 and @xmath188 is essentially non - zero . hence , @xmath189 and the maximum position for @xmath190 is shifted to larger values as compared to that for @xmath191 , as observed in fig . [ l123_dist ] ( a ) . due to these two requirements , the distributions @xmath190 and @xmath191 can not be reduced to the same master curve by the scaling transformation only . we find , however , that the @xmath183 distribution can be matched with the others by applying a shift @xmath192 , which is not surprising , as it turns the shifted value into some approximation of the @xmath5 ( this is not shown in the figure for a sake of brevity ) . nevertheless , one may report relative proximity of the probability distributions for the scaled properties @xmath193 and @xmath194 indicating strong isotropicity of the polymer chain along with its eigenvalues . indeed , not only the mean values for all eigenvalues scale by the same law [ fig . [ scaling_l123rgsq ] ( a ) ] , but also their probability distributions do so to great extent . probability distribution @xmath162 and the results of fits according to eq . ( [ gauss ] ) ( marked as [ g ] ) , eq . ( [ lhuillier ] ) ( marked as [ l ] ) and eq . ( [ lhuillier2 ] ) ( marked as [ l@xmath137 ) . the parameters of fits are shown in a figure.,width=264 ] the prediction of the asymptotics for the probability distributions of the shape characteristics similar to the way it is done by lhuillier @xcite for @xmath5 faces severe difficulties . indeed , the interval of small @xmath5 is associated with a collapsed chain , whereas that of large @xmath5 to highly stretched conformations . this does not hold for the asphericity @xmath60 , where small ( large ) values of @xmath60 contain contributions from the whole range of conformations being spherically symmetric ( asymmetric ) . as was shown in ref . @xcite , the form of the probability distributions @xmath162 and @xmath164 differ for the cases of saw and rw . therefore , reparametrisation of the chi - squared expressions valid for a rw do not reproduce well the respective distributions for the saw case . in this paper we chose another route : reparametrisation of the lhuillier - like expression ( [ lhuillier ] ) , heuristically suggested for the gyration radius distribution , and its extension for the distribution of shape characteristics . similarly to the cases of the @xmath4 and @xmath5 probability distributions , we build a cumulative probability distribution histogram for each shape characteristic based on the simulation data for @xmath195 , @xmath196 , @xmath197 , @xmath198 and @xmath199 . as was shown in figs . [ scaling_l123rgsq]-[scaling_gr12r13 ] , within this interval of @xmath1 the scaling laws follow sufficiently well and the probability distributions are found to overlap . the study of such a cumulative histogram essentially enhances the statistics of the analysed data . the form of the probability distribution @xmath162 for the asphericity , shown in fig . [ a_dist ] , indicates a relatively low asymmetry , therefore we attempt fits using both functional forms ( [ gauss ] ) and ( [ lhuillier ] ) . this yield the following exponents : @xmath200 , @xmath201 and @xmath202 and the most probable values @xmath203 and @xmath204 for @xmath205 and @xmath206 , respectively . the fit _ via _ lhuillier - like form follows the shape of @xmath162 more closely . an even better approximation can be achieved by using the extended lhuillier - like form @xmath207,\;\;\ ; \bar{x}=\big(x_1^{\epsilon_1}\ , x_2^{\epsilon_2}\,\frac{\epsilon_1}{\epsilon_2}\big)^{\frac{1}{\epsilon_1+\epsilon_2}}.\ ] ] in this case the exponents are : @xmath208 and @xmath209 and the most probable value for the asphericity is @xmath210 . this is very close to the one obtained with the ( [ lhuillier ] ) form . the most probable values @xmath127 are added to tab . [ tab_aver ] and one notes that the value obtained from the fit ( [ gauss ] ) is close to the average value @xmath211 , whereas both values obtained _ via _ fits ( [ lhuillier ] ) and ( [ lhuillier2 ] ) are closer to the asphericity of a rw . probability distribution @xmath164 and the results of a fit according to eq . ( [ sfit ] ) . the parameters of a fit are shown in a figure.,width=226 ] the prolateness @xmath61 , as already mentioned above , distinguish es between the oblate ( @xmath212 ) and prolate ( @xmath213 ) conformations of a chain . the shape of its probability distribution , @xmath164 , is also markedly different in these two intervals of @xmath61 values , see fig . [ s_dist ] . we were unable to fit it by a single analytic form but instead opted for two separate fits , at @xmath212 and @xmath213 @xmath214 $ , & $ s>0$. \end{tabular } \right.\ ] ] the analytic form used for @xmath213 can be attributed to a limit case of both eq . ( [ gauss ] ) ( at @xmath215 ) and of eq . ( [ lhuillier2 ] ) ( at @xmath216 ) , therefore this fit is marked as [ g , l@xmath217 or l ] in fig . [ s_dist ] . one can not define the most probable value @xmath218 for the prolateness from a fit ( [ sfit ] ) . probability distribution @xmath219 and the results of fits according to eq . ( [ gauss ] ) ( marked as [ g ] ) and eq . ( [ lhuillier ] ) ( marked as [ l ] ) . the parameters of fits are shown in the figure.,width=264 ] the probability distribution @xmath219 for the size ratio @xmath62 is shown in fig . [ g_dist ] . it is fitted similarly to the @xmath162 distribution , _ via _ analytic forms ( [ gauss ] ) and ( [ lhuillier ] ) . this yields the respective exponents @xmath220 , @xmath221 and @xmath222 . the most probable values are @xmath223 and @xmath224 , respectively . one should note that other approaches evaluate the @xmath113 value only which is equal to @xmath225 for the rw and is found to be larger that @xmath225 for the saw ( see , tab . [ tab_aver ] ) . it is evident from fig . [ scaling_gr12r13 ] that reliable estimates for both @xmath113 and @xmath116 require the use of a polymer chain longer than @xmath226 ( maximal length used in this study ) and , therefore , can not be provided within this study . therefore , the most probable values @xmath128 obtained here do not have any counterparts to be compared with and can serve as an indication of high asymmetry of the @xmath219 distribution only . ( a ) probability distribution @xmath227 and the results of a fit according to eq . ( [ lhuillier ] ) with fixed @xmath228 ( marked as [ l ] ) . ( b ) the same for @xmath229 . the parameters of fits are shown in the figure.,title="fig:",width=226 ] ( a ) probability distribution @xmath227 and the results of a fit according to eq . ( [ lhuillier ] ) with fixed @xmath228 ( marked as [ l ] ) . ( b ) the same for @xmath229 . the parameters of fits are shown in the figure.,title="fig:",width=226 ] the probability distributions @xmath227 and @xmath229 appear to be of a similar , highly asymmetric form , see fig . [ r_dist ] ( a ) and ( b ) . both were preliminarily fitted by the general expression ( [ lhuillier ] ) yielding the exponents @xmath230 and @xmath231 close to @xmath232 . therefore , in the final fits we set @xmath228 and obtain the following estimates for the most probable values : @xmath233 and @xmath234 , the fits are shown in fig . [ r_dist ] . these values differ about @xmath235 times from their counterparts @xmath117 and @xmath118 ( see , tab . [ tab_aver ] ) due to high asymmetry of respective distributions . here we have analysed the shape characteristics of a coarse - grained polymer chain in a good solvent using dpd simulations . one of the aims was to check the universality of the properties not directly connected to the scaling power laws . in this respect we here evaluated the shape characteristics which can be seen as counterparts of the universal critical amplitude ratios in critical phenomena . as far as the simulations employ an off - lattice model for the polymer chain and an explicit solvent , these are more computationally intensive than their lattice monte carlo counterparts and can not match the latter in numerical accuracy . however , using this simulation technique , one bridges the gap to a large number of studies targeted on at self - assembly of various amphi- and polyphilic systems and to the flow - driven phenomena on the nanoscale . hence , according to our aim , we rather focus on a qualitative analysis of the mean and most probable values for the shape characteristics , as well as on the form of their probability distributions . therefore , no rigorous analysis of the numerical errors was performed and these are not displayed in the figures or in the table of results . the origin our analysis is a study of the scaling properties of the gyration tensor components . we find that , for chain lengths @xmath236 , not only the end - to - end distance and gyration radius , but also all three eigenvalues of the gyration tensor scale with the same power law . the flory exponent @xmath3 obtained for each of these properties independently , within the accuracy of the simulations , is close to the best theoretical estimate @xmath102 . based on the analytic expressions , the shape characteristics of the chain are expected to be independent of @xmath1 in the same interval of chain lengths . this has been checked by plotting each of them against @xmath1 . the mean values for the shape characteristics were evaluated by averaging their respective values obtained for chain lengths of @xmath195 , @xmath196 , @xmath197 , @xmath198 and @xmath199 . the mean values for the asphericity and prolateness are found to compare well with the respective results obtained by means of other approaches , whereas the shape anisotropy requires simulations of longer chains . the averaging procedure for these properties , however , is found to be rather ambiguous , as the probability distributions for all shape characteristics are wide and many are highly asymmetric . therefore , alongside with the mean values , one may consider looking at the most probable values of each characteristic as well . this can be performed by fitting the probability distribution for each characteristic by certain analytic expressions and determining the most probable value based on the positions of its maxima . here , we started from the analysis of the probability distributions for the reduced end - to - end distance and the gyration radius . for the former case , the known analytic asymptotics of cloiseaux - de gennes were reproduced , whereas for the latter the lhuillier analytic expression is found to be very accurate yielding another independent estimate for the flory exponent @xmath3 . the analytic expressions in both cases are based on an analysis of the conformation statistics , which is more difficult to apply to the shape characteristics . for the latter , we suggest heuristic analytic expressions based on the lhuillier form and , for the symmetric distributions , on a generalised gaussian distribution . this leads to simple analytical expressions with the coefficients and exponents found from fitting the simulation data . this allows us to complement the mean values for the shape properties by their most probable values found as the maximal positions of the respective distributions . the study can be interpreted as yet another validation on the appropriateness of the use of soft coarse - grained potentials to describe the self - avoiding polymer chain . not only the scaling properties of the end - to - end distance and gyration radius of such chains are in good agreement with the theoretical data , but also the mean values of the shape characteristics and their probability distributions are found to closely mimic the results of more accurate lattice monte carlo simulations . we are thankful to victoria blavatska for useful discussions . this work was supported in part by fp7 eu irses projects no . @xmath237 statistical physics in diverse realizations " , no . dynamics of and in complex systems " , no . @xmath239 structure and evolution of complex systems with applications in physics and life sciences " , and by the doctoral college for the statistical physics of complex systems , leipzig - lorraine - lviv - coventry @xmath240 . zifferer g 1999 _ macromolecular theory and simulations _ 8 * 433462 http://dx.doi.org/10.1002/(sici)1521-3919(19990901)8:5<433::aid-mats433>3.0.co;2-c[http://dx.doi.org/10.1002/(sici)1521-3919(19990901)8:5<433::aid-mats433>3.0.co;2-c ]	in this paper we study the shape characteristics of a polymer chain in a good solvent using a mesoscopic level of modelling . the dissipative particle dynamics simulations are performed in the @xmath0 space at a range of chain lengths @xmath1 . the scaling laws for the end - to - end distance and gyration radius are examined first and found to hold for @xmath2 yielding reasonably accurate value for the flory exponent @xmath3 . within the same interval of chain lengths , the asphericity , prolateness , size ratio and other shape characteristics of the chain are found to become independent of @xmath1 . their mean values are found to agree reasonably well with the respective theoretical results and lattice monte carlo simulations . broad probability distributions for the shape characteristics are found resembling in form the results of lattice monte carlo simulations . by means of analytic fitting of these distributions the most probable values for the shape characteristics are found to supplement their mean values . _ keywords _ : polymer , scaling , dissipative particle dynamics
in this section we will compare the scpt ( see , e.g. , @xcite ) and ept @xcite . to make this concrete , we mainly focus on the temperature dependence of the physical quantities of the fermi - pasta - ulam @xmath0 ( fpu-@xmath0 ) model , which is a classic example to study heat conduction . the hamiltonian of the fpu-@xmath0 model is given by @xmath13 within the scp approximation , the hamiltonian is approximated by a trial hamiltonian @xmath14 where the effective harmonic coupling constant @xmath15 is given by eq . in the appendix with the boltzmann constant @xmath16 . note that @xmath15 is temperature dependent , which stems from the existence of nonlinearity in the system . the dispersion relation of effective phonons corresponding to the effective hamiltonian can be written in the form @xmath17 it is then straightforward to calculate the dimensionless nonlinearity @xmath10 and the specific heat @xmath18 from gaussian averages of scpt , which yields @xmath19 @xmath20 the derivation of the effective phonon spectrum is based on the generalized equipartition theorem , @xmath21 where the bracket @xmath22 stands for a thermal average in the canonical ensemble and @xmath23 denotes the fourier transform of the coordinate @xmath24 . the next step consists of an approximative transformation of the right hand side of eq . into a more compact form ( see @xcite for details ) , @xmath25 specifically , the effective phonon spectrum @xmath26 for the fpu-@xmath0 model reads as @xmath27 here @xmath28 is given by @xmath29 } , \ ] ] where @xmath30 and @xmath31 are the modified bessel functions of the second kind and @xmath32 . the parameters for the nonlinearity @xmath33 and the specific heat @xmath34 that follow from ept can be expressed in the form @xmath35 + 2\lambda t y_{1/4}(x ) } { k^2[y_{3/4}(x)-y_{1/4}(x)]+2\lambda t y_{1/4}(x)},\ ] ] @xmath36 in fig . [ fcst ] , we plot the effective sound speed @xmath37 ( and @xmath38 ) as a function of the temperature . the high temperature behavior gives @xmath39 , which was already reported in ref . we also compare @xmath10 and @xmath18 calculated from scpt and ept in fig . [ fepst ] and fig . [ ct ] , respectively . one can notice that scpt and ept yield practically the same temperature dependence over seven orders of magnitude . effective sound speed @xmath40 as a function of temperature . at high temperature regime , @xmath41 . here @xmath42 , @xmath43.,width=5 ] temperature dependence of the dimensionless nonlinearity @xmath10 . at low temperature regime @xmath44 , while at high temperature limit @xmath45 . here @xmath42 , @xmath43.,width=5 ] temperature dependence of specific heat . @xmath46 at low temperature regime ( harmonic limit ) and @xmath18 approaches the lower limit @xmath47 at high temperature regime . here @xmath42 , @xmath43.,width=5 ] we will now apply these two effective phonon approaches to calculate the thermal conductivity from the debye formula . for the sake of simplicity , we will only present the derivation from scpt . the phonon group velocity within scpt is given by @xmath48 according to the assumption for the relaxation time , the mean free path reads as @xmath49 substituting eq . and eq . into eq . , the temperature dependence of the thermal conductivity can be given by @xcite @xmath50 the calculation of @xmath2 from ept follows along similar lines . by replacing in eq . ( [ kappa ] ) the specific heat , the group non - linearity and the non - linearity by their counterpart deduced from ept , that is eq . , and , one obtains an alternative expression for the thermal conductivity ( [ kappa ] ) . note that we only consider the temperature dependence of @xmath2 here as in ref . nevertheless , it should be emphasized that @xmath2 is both temperature and size dependent . @xmath2 diverges in the thermodynamic limit due to the divergence of summation with respect to @xmath6 in the acoustic regime , which is not the concern of this study . the analytical prediction and the simulation result for the temperature dependence of @xmath2 are compared in fig . [ ktb1 ] and fig . [ ktb01 ] for @xmath43 and @xmath51 , respectively . one can notice that the power law behavior observed in ref . @xcite , i.e. , @xmath52 at low temperature regime and @xmath53 at high temperature regime are exactly reproduced . as a comparison , non - equilibrium molecular dynamics simulations were performed by applying langevin heat baths at the two ends of the chain @xcite . in order to compute the thermal conductivity , a finite temperature difference with @xmath54 deviation from the average temperature was consistently used . it is clearly seen that the analytical calculations , rescaled by a constant factor that is implicit in the debye formula , give good agreements with the simulation results in a large range of temperature . temperature dependence of thermal conductivity . here @xmath42 , @xmath43 , and system size @xmath55 for simulation . analytical results ( solid line and dot line ) are rescaled ( divided by a constant ) , which is consistent with simulation result in whole temperature range . the simulation result shows the temperature for the minimum of @xmath2 at @xmath56 , in agreement with eq . .,width=5 ] temperature dependence of thermal conductivity . here @xmath42 , @xmath51 , and @xmath55 for simulation . like fig . [ ktb1 ] , analytical results are rescaled . the simulation result shows @xmath57 , in agreement with eq . .,width=5 ] both fig . [ ktb1 ] and fig . [ ktb01 ] show the existence of a minimum of @xmath2 at a specific temperature @xmath58 . by computing the derivative of the thermal conductivity with respect to @xmath59 , @xmath60 one easily obtains the turning point in the simple form @xmath61 here we neglect the temperature dependence of the classical specific heat @xmath18 , which has a very weak dependence on the temperature . the dimensionless constant @xmath62 is then given by @xmath63 if one takes into account the specific heat given by , the behavior of @xmath64 survives while @xmath62 should be replaced approximately by @xmath65 . the turning point @xmath58 denotes the transition of @xmath2 from decreasing behavior to increment behavior with increasing temperature . . can be understood by scaling the hamiltonian as done in ref . let [ x_scaling ] @xmath66 one obtains @xmath67 where the dimensionless hamiltonian @xmath68 reads as @xmath69 eq . leads to @xmath70 eq . means that the temperature dependence of physical quantities of the fpu-@xmath0 model can be rescaled according to the temperature transformation in eq . , which can be verified in fig . [ scaling ] . scaling of thermal conductivity . here we simply apply coordinate transformation @xmath71 to fig . [ ktb01 ] and then merge it with fig . one can see the overlap of the two figures , which indicates the scaling of thermal conductivity.,width=5 ] note that the nonlinearity @xmath72 dependence of @xmath2 is similar with the temperature dependence shown in fig . [ ktb1 ] , fig . [ ktb01 ] and fig . [ scaling ] . a critical nonlinearity @xmath73 corresponding to the minimum of @xmath2 can also be similarly obtained as @xmath74 eq . and eq . shows that the temperature @xmath59 and the nonlinearity @xmath72 are inversely equivalent . finally , it is interesting to note that @xmath75 which is the well - known golden ratio . due to the particular property of the golden ratio , the ratio of nonlinear potential and harmonic potential @xmath76 is also the golden ratio . from the experimental point of view , the thermal conductivity of real materials should vanish at zero temperature . this behaviour is inexistent within the classical description of heat conduction as shown above , since the latter leads to a divergent thermal conductivity at zero temperature . the disagreement between the classical description and experimental curves results from the inability of the classical physics to take into account the freezing of phonon modes , which is a pure quantum effect . we would like to emphasize that the consideration of the low temperature regime is not of pure theoretical interest . generally , the debye temperature of many solid systems is indeed well above ambient temperature . to name but a few , @xmath77 k in stoichiometric cab@xmath78 and @xmath79 k for vacancy doped eub@xmath78 @xcite . in carbon nanotubes , the debye temperature may even exceed @xmath80 k @xcite . one thus should take into account quantum effects , namely the partial thermalization of phonon modes even far above the room temperature for these materials . in the following , we will treat the debye equation for the fpu-@xmath0 model semi - classically and show that the vanishing behaviour of @xmath2 manifests itself if one correctly takes into account the bose - einstein statistics . at this point , we should recall that the derivation of ept is based on the classical equipartition theorem ( see eq . ) . it is thus impossible to extent the effective phonon approach to the quantum regime , which makes scpt more adequate since the latter can be derived from a purely quantum approach . the derivation of the quantum self - consistent phonon theory ( qscpt ) is presented in the appendix . to avoid the divergence of eq . due to the goldstone mode ( @xmath81 ) , a small quadratic on - site potential @xmath82 is included in the hamiltonian . here the constant @xmath83 is fixed to its lower - boundary ( @xmath84 ) so that its decrease below this limit does nt change the value of @xmath2 . a strong point of qscpt is that the approach unambiguously provides a mode decomposition for the specific heat and the phonon velocity , which allows one to consider in a rigorous way the discrete mode summation of the debye equation . the specific heat per mode @xmath3 is obtained from . the phonon velocity @xmath4 follows from the derivative of the pseudo - phonon spectrum and according to eq . , the mean free path is given by @xmath85 where @xmath86 . for this model , the phonon frequency that should be numerically solved with eq . is @xmath87 where the integer @xmath88 $ ] and the phonon mode @xmath89 . the non - linearity parameter is given by @xmath90 fig . [ qct ] gives the specific heat @xmath91 as a function of the temperature . at the low temperature regime , @xmath18 increases linearly with increasing temperature since the high energy modes are gradually thermalized . this is the well - known partial thermalization effect . the agreement with the rigorous result for the harmonic chain ( @xmath92 ) at low temperature regime shows that the harmonic quantum fluctuations play the dominant role . the quantum effect , as expected , is negligible above the debye temperature @xmath93 and the system behaves like the classical one . specific heat as a function of temperature for ( a ) @xmath43 ; ( b ) @xmath51 . the harmonic potential is given by @xmath94 . @xmath42 and @xmath95 for all cases . at low temperature regime , @xmath96.,width=5 ] by collecting all the quantities in the debye equation , we finally obtain the temperature dependence of the thermal conductivity from the quantum approach , which we compare in fig . [ qkt ] with its classical counterpart . when the temperature is high enough , qscpt reproduces the classical behavior as shown in fig . [ ktb1 ] and fig . [ ktb01 ] , where the mean free path and the specific heat are nearly constant and the temperature dependence of @xmath2 mainly follows that of the sound speed @xmath39 . the classical and quantum results agree up to the debye temperature . the deviation of the quantum result takes place around @xmath97 , which is followed by a peak at temperature @xmath98 , then the thermal conductivity drops to zero . @xmath99 here indicates the lower bound of the temperature range over which _ umklapp _ processes @xcite yield the dominant contribution to thermal conductivity . the existence of this peak results from a competition between the increasing behavior of the mean free path and the strongly decreasing tendency of the specific heat . the latter property is a consequence of the partial thermalization of high frequency phonons , which is a pure quantum effect explaining the failure of classical approaches below the debye temperature . at low temperatures , the mean free path and the sound speed are nearly constant in the quantum case . the difference for the temperature dependence of @xmath2 is mainly due to the difference of the specific heat . our result based on qscpt , characterized by the existence of the _ umklapp _ peak , is qualitatively consistent with the experimental studies ( see , e.g. , @xcite ) . finally , one should note that the scaling behavior is not applicable for quantum case . temperature dependence of thermal conductivity for ( a ) @xmath43 ; ( b ) @xmath51 . @xmath42 and @xmath84 for both cases.,width=5 ] in summary , scpt and ept are compared in the present study . the basic difference between scpt and ept lies in the different way to calculate the statistical average of physical quantities . specifically , the canonical average is applied in ept but gaussian average is applied in scpt . we show that both scpt and ept give the same behavior for the thermodynamic quantities of a classical system , which indicates their equivalence . it seems that gaussian average is preferable for analytical derivation , especially for models with an on - site potential , since the thermal averages should be computed in this case from numerical transfer matrix calculations . the additional gain is that scpt can be extended to study the quantum system , as shown in section ii . note that ept is based on the assumption of the equipartition theorem , whose validity in the quantum regime is broken . thus qscpt offers a simple way to explore for a given hamiltonian model the low temperature physics related to the bose - einstein statistics . we studied the temperature dependence of the thermal conductivity of the fpu-@xmath0 model for both the classical and quantum case . the temperature for the minimum of @xmath100 was determined from the classical scpt , which clearly shows the scaling behavior of thermal conductivity for this non - linear model . we also showed that non - equilibrium molecular dynamics simulations are in good agreement with the analytical results . at the low temperature regime , the semiclassical treatment of the debye equation reproduced the _ peak , a well - known characteristic observed in experimental studies . this work was supported in part by grants from the hong kong research grants council ( rgc ) and the hong kong baptist university . in this appendix , we present a simple derivation of the first order self - consistent phonon theory for 1d hamiltonian systems . the presented approach is similar to the derivation of the variational path integral method ( see @xcite ) . in the path integral representation of quantum statistical mechanics , the @xmath101-body partition function in the canonical ensemble can be expressed as a path integral over periodic trajectories , that is @xmath102},\ ] ] where @xmath103\right)\ ] ] is the euclidean action and the first and second terms in the integral correspond to the total kinetic and potential energies . it is well known that an exact evaluation of the @xmath101-body path integral ( [ a1 ] ) for general anharmonic potentials is impossible . the idea of scpt is basically approximating the original euclidean action with a trial action that allows an exact evaluation of the trace ( [ a1 ] ) . since we deal in this work exclusively with 1d hamiltonian models with closest neighbor interactions of the form @xmath104 where @xmath105 , an optimal choice of the trial hamiltonian is that of a coupled harmonic oscillator chain , @xmath106 the trial parameters @xmath107 , @xmath108 and @xmath109 are to be deduced by minimizing the right hand - side of the feynman - jensen inequality , @xmath110 where @xmath111 . the trial partition function given by @xmath112/\hbar},\ ] ] can be easily computed by first performing a fourier decomposition of the periodic paths , @xmath113 where @xmath114 stand for matsubara frequencies and the measure of integration of eq . ( [ a5 ] ) has the form @xmath115 in the last expression as well as in eq . ( [ a6 ] ) , index @xmath6 run over oscillators and @xmath116 over fourier components . we then substitute the expansion ( [ a6 ] ) into ( [ a3 ] ) and obtain the trial action by performing the integration over time @xmath12 , @xmath117 the trial action @xmath118 that follows is a quadratic function of the fourier components @xmath119 , that is @xmath120 the next step consists of the trivial integration in eq . ( [ a5 ] ) over fourier components , which yields @xmath121 for @xmath122 and @xmath123 otherwise . in the last two expressions which give the zero - th order free energy @xmath124 , we have defined the pseudo - phonon frequencies in the form @xmath125 and an additional parameter @xmath126 the calculation of the first order correction to free energy @xmath127}\left\{\sum_{k=1}^n \left[v(x_k)+w(\delta x_k)\right]-h_0\right\}\ ] ] proceeds in a similar way . the usual trick consists in expanding the on - site and inter - site potential in fourier basis , @xmath128 then we evaluate the average potentials per particle @xmath129 and @xmath129 in ( [ a13 ] ) by integrating over fourier components and finally invert the fourier transforms of eq . ( [ a14 ] ) . consequently , the average value of the potential energies can be expressed in the form of smeared - out potentials , @xmath130 where we have defined two parameters @xmath131 and @xmath132 which correspond to lattice displacement and two - point correlation function , that is @xmath133 and @xmath134 the gaussian smearing ( [ a15 ] ) is a key characteristic of variational methods . the first order free energy per particle can be finally expressed as @xmath135 by minimizing this expression with respect to @xmath136 , one obtains the optimal pseudo - phonon frequency @xmath137 on the other hand , the variation of eq . ( [ a18 ] ) with respects to @xmath138 yields @xmath139 note that @xmath140 for pair on site potentials ( @xmath141 ) . using ( [ a15 ] ) , ( [ a16 ] ) and ( [ a19 ] ) in eq . ( [ a18 ] ) , we can further simplify the first order free energy in the form @xmath142 the specific heat per mode can be then defined by @xmath143 one can easily obtain the classical limit of by letting @xmath144 . the final result is @xmath145 where @xmath131 and @xmath132 in eq . ( [ a21 ] ) are replaced by their classical counterpart , @xmath146 and @xmath147	we compare two effective phonon theories , which have both been applied recently to study heat conduction in anharmonic lattices . in particular , we study the temperature dependence of the thermal conductivity of fermi - pasta - ulam @xmath0 model via the debye formula , showing the equivalence of both approaches . the temperature for the minimum of the thermal conductivity and the corresponding scaling behavior are analytically calculated , which agree well with the result obtained from non - equilibrium simulations . we also give quantum corrections for the thermal conductivity from quantum self - consistent phonon theory . the vanishing behavior at low temperature regime and the existence of an _ umklapp _ peak are qualitatively consistent with experimental studies . the study of heat conduction is very important from both theoretical and experimental point of view . a traditional phenomenological approach to understand the thermal properties in solids is the debye formula given by @xmath1 where @xmath2 is the thermal conductivity , @xmath3 , @xmath4 , @xmath5 are the specific heat , the phonon group velocity and the phonon mean free path of mode @xmath6 , respectively . in spite of its successfulness for qualitative explanation of heat conduction in dielectrics , quantitative predictions are hard to make from a microscopic viewpoint . recently , an increasing study of heat conduction in low dimensional hamiltonian models may shed light on its microscopic understanding @xcite . however , only a few integrable models can be solved rigourously @xcite . generally one has to rely on numerical simulations for non - integrable models . thus it would be worthy to revisit the traditional kinetic approach by incorporating some microscopic consideration for the low dimensional non - integrable lattice systems . according to the debye formula , the thermal transport process in a nonlinear lattice is intrinsically relative to its dispersion relation and relaxation of normal modes ( phonons ) . the existence of nonlinearity makes the definition of phonon delicate , and it is even harder in this case to quantify phonon transport from a first - principle way . to surmount the difficulties due to nonlinearity , the concept of `` effective phonons '' has been recently introduced to study heat conduction in dynamical models @xcite . the basic idea consists in incorporating the non - linearity into normal modes by renormalizing the harmonic frequency spectrum . in the serial studies @xcite , the authors apply the so - called effective phonon theory ( ept ) to study heat conduction within the kinetic framework . in ref . @xcite , the authors study heat conduction through a lattice consisting of two weakly coupled nonlinear segments via the self - consistent phonon theory ( scpt ) . it is thus interesting to compare ept and scpt since they have both been applied to the field of heat conduction . in the present study , we will give a detailed comparison via the debye formula as done in @xcite . our result shows the equivalence of scpt and ept in a large range of temperature . considering the failure of the classical description at low temperature regime , we also compute quantum corrections to the thermal conductivity by extending our study to quantum regime , which gives qualitatively consistent results with experimental studies . it should be emphasized that both ept and scpt can not give the microscopic definition of the relaxation time @xmath7 for phonons of mode @xmath6 . a traditional perturbative way consists in studying the single mode relaxation time based on three- and four- phonon processes @xcite . great efforts have been devoted to obtain the relaxation time of the heat current correlations @xcite , which in general might be related to the relaxation time of phonons . however , the relation was so far unclear . for an anharmonic chain , a simple but physically appealing assumption @xmath8 has been proposed in ref . @xcite , where @xmath9 is the phono frequency of mode @xmath6 . the dimensionless nonlinearity @xmath10 is defined as the ratio between the average of the anharmonic potential energy and the total potential energy : @xmath11 one can see that , when @xmath10 vanishes , @xmath12 approaches to infinity , leading to the expected divergence of the thermal conductivity . in the following , we will apply the assumption of eq . to eq . in order to study the thermal conduction of an anharmonic chain , which gives surprising agreement with non - equilibrium simulations .
nonthermal hard x - ray ( hxr ) radiation has been detected for the first time in the coma cluster by ( fusco - femiano 1999 ) and _ rxte _ ( rephaeli , gruber & blanco 1999 ) , while marginal evidence is reported for a2199 ( kaastra 1999 ) . these observations are only first steps towards assessing the general existence of this new component in the x - ray spectra of clusters of galaxies . the search for nonthermal emission in more clusters is of high importance as it will allow to derive additional informations on the physical conditions of the intracluster medium ( icm ) environment , which can not be obtained by studying the thermal plasma emission only . various interpretations of the hxr emission have been presented since its discovery in the coma cluster spectrum . the most direct explanation is inverse compton ( ic ) scattering of cosmic microwave background ( cmb ) photons by the relativistic electrons responsible of the extended radio emission present in the central region of coma ( willson 1970 ) . the combined radio synchrotron and ic hxr fluxes ( e.g. , rephaeli 1979 ) allow to estimate a volume - averaged intracluster magnetic field of @xmath4 ( fusco - femiano 1999 ) . one of the problems with the ic model is that this value of the magnetic field in the icm seems to be at odd with the value determined from faraday rotation of polarized radiation toward the head tail radio galaxy ngc4869 that gives a line - of - sight @xmath5 ( feretti 1995 ) , and with the equipartition value in the radio halo , which is @xmath6 ( giovannini 1993 ) . we note , however , that feretti ( 1995 ) also inferred the existence of a weaker and larger scale magnetic field component in the range of @xmath7 , and therefore the @xmath8 field could be local . a low average magnetic field is also consistent with the model developed by brunetti ( 1999 ) , which predicts a magnetic field strenght decreasing with the distance from the cluster centre . an alternative explanation is nonthermal bremsstrahlung ( ntb ) emission from suprathermal electrons currently accelerated at energies greater than @xmath010 kev by shocks or turbulence ( kaastra 1998 ; ensslin , lieu , & biermann 1999 ; sarazin & kempner 1999 ) . another and more trivial possibility is that the hxr radiation is due to a hard x - ray source present in the external regions of the field of view of the pds ( fwhm=@xmath9 , hexagonal ) , as for example a highly obscured seyfert 2 galaxy like the circinus galaxy ( matt 1999 ) . in the central region ( @xmath10 in radius ) , the mecs image does not show evidence of this kind of sources ( fusco - femiano 1999 ) . hovewer , the detection of a hard nonthermal component in other clusters should strongly reduce the probability of this last interpretation . in this letter we present the results of a long observation of a2256 , exploiting the unique capabilities of the pds , onboard , to search for hxr emission ( frontera 1997 ) . the cluster was also observed with the mecs , an imaging instrument working in the 1.5 - 10 kev energy range ( boella 1997 ) . the galaxy cluster a2256 is similar to the coma cluster in many x - ray properties , as luminosity and presence of substructures . the pspc observations showed that a2256 is a double x - ray cluster ( briel 1991 ) , suggesting that a subcluster may be merging with a larger cluster , although there is no strong evidence in the temperature map in favour of an advanced merger ( markevitch & vikhlinin 1997 ) , as it is for coma . the average gas temperature is @xmath07 kev , as measured by several x - ray instruments ( david 1993 ; hatsukade 1989 ; markevitch & vikhlinin 1997 ; henriksen 1999 ) . both clusters show a radio halo in the central and periferal regions . however , the radio emission from a2256 is notably complex . the region around the cluster centre is occupied by an unusual concentration of radio galaxies : at least five discrete sources have been identified with cluster galaxies , but there also two extended emission regions which have linear sizes @xmath111 mpc ( bridle & formalont 1976 ; bridle 1979 ; rottgering 1994 ) . throughout the letter we assume a hubble constant of @xmath12 and @xmath13 , so that an angular distance of @xmath14 corresponds to 92 kpc ( @xmath15 ; struble & rood 1991 ) . quoted confidence intervals are at @xmath16 level , if not otherwise specified . the total effective exposure time was @xmath17 sec for the mecs and @xmath18 sec for the pds in the two observations of february 1998 and february 1999 . the observed count rate for a2256 was [email protected] cts / s for the 2 mecs units and [email protected] cts / s for the pds instrument . since the source is rather faint in the pds band ( approximately 1.5 mcrab in 15 - 150 kev ) a careful check of the background subtraction must be performed . the background sampling was performed using the default rocking law of the two pds collimators that samples on , + off , on , -off fields for each collimator with a dwell time of 96 " ( frontera et al . 1997 ) . when one collimator is pointing on source , the other collimator is pointing toward one of the two off positions . we used the standard procedure to obtain pds spectra ( dal fiume et al . 1997 ) , which consists in extracting one accumulated spectrum for each unit for each collimator position . we then checked the two independently accumulated background spectra in the two different + /-off sky directions , offset by 210 with respect to the on - axis pointing direction . the comparison between the two accumulated backgrounds ( [ + off ] - [ -off ] ) shows a difference with a marginal excess below 30 kev in the [ + off ] pointing . this excess is much lower than the signal from the source , but it must not be neglected . the total excess in the first two equalized energy channels ( 15 - 33.5 kev ) is @xmath20 , i.e. approximately 2@xmath1 . this concentration in only the lowest energy channels implies that the excess is likely due to contamination by a point source rather than to a statistical fluctuation . the total source spectrum was therefore obtained using only the uncontaminated background accumulated pointing at the [ -off ] field . hovewer , in section 3 we report the confidence level of the nonthermal emission in excess of the thermal one considering the average of the two background measurements . the background level of the pds is the lowest obtained thus far with high - energy instruments on board satellites thanks to the equatorial orbit and is very stable again thanks to the favorable orbit . no modeling of the time variation of the background is required . mecs data preparation and linearization was performed using the saxdas package under ftools environment . we have extracted a mecs spectrum from a circular region of 8@xmath21 radius ( corresponding to about 0.8 mpc ) centered on the primary emission peak . from the rosat pspc radial profile , we estimate that about 70% of the total cluster emission falls within this radius . the background subtraction has been performed using spectra extracted from blank sky event files in the same region of the detector as the source . a numerical relative normalization factor among the two instruments has been included in the fitting procedure ( see next section ) to account for : a ) the fact that the mecs spectrum includes emission out to @xmath00.8 mpc from the x - ray peak , while the pds field of view ( 1.3 degrees fwhm ) covers the entire emission from the cluster ; b ) the slight mismatch in the absolute flux calibration of the mecs and pds response matrices employed ( september 1997 release ; fiore , guainazzi & grandi 1999 ) ; c ) the vignetting in the pds instrument , ( the mecs vignetting is included in the response matrix ) . the estimated normalization factor is @xmath01.1 . in the fitting procedure we allow this factor to vary within 15% from the above value to account for the uncertainty in this parameter . the spectral analysis of the mecs data alone , in the energy range 2 - 9.7 kev and in the central @xmath00.8 mpc region , gives a gas temperature of @xmath22 kev ( @xmath23=154.5 for 162 degrees of freedom ; hereafter dof ) , using an optically thin thermal emission model ( mekal code on the xspec package ) , absorbed by a galactic line of sight equivalent hydrogen column density , @xmath24 , of 4.01@xmath25 . this value of the temperature is consistent with the gis measurement ( 6.78 - 7.44 kev ; henriksen 1999 ) , and with the values obtained by previous observations : the mpc ( 6.7 - 8.1 kev ; david 1993 ) and ( 7.32 - 7.70 kev ; hatsukade 1989 ) . also the flux of @xmath26 in the 2 - 10 kev energy range is consistent with the previous measurements . the iron abundance is @xmath27 , in agreement with the results ( markevitch & vikhlinin 1997 ) . the analysis of the pds data with a thermal bresstrahlung component gives a temperature of @xmath030 kev . fitting the data with two thermal components , one of these at the fixed temperature of 7.4 kev , we obtain a temperature greater than @xmath090 kev for the second component . these unrealistic high values for the gas temperature obtained in both the fits are interpreted as a strong indication that the detected hard excess is due to a nonthermal mechanism . figure 1 shows the simultaneous fit to the mecs and pds data with a thermal component at the temperature of [email protected] kev and a normalization factor of @xmath01.2 for the two data sets . the @xmath23 is 180.5 for 167 dof . hard x - ray radiation at energies greater than @xmath020 kev is in excess with respect to the thermal component at a level of @xmath28 and this value is rather stable against variation of the normalization factor . it results slightly lower ( @xmath29 ) considering the average of the two background measurements . besides , also fitting the pds data alone with a thermal component at the fixed temperature of 7.47 kev we obtain an excess at a level of 4.3@xmath1 . if we introduce a second nonthermal component , modeled as a power law , we obtain the fit shown in figure 2 . the @xmath23 is 156.6 for 165 dof . the improvement with respect to the previous model is significant at more than the 99.99% confidence level , according to the f - test . the confidence contours of the parameters @xmath30 and photon spectral index ( @xmath31 ) show that , at 90% confidence level , the temperature is well determined , 6.8 - 7.5 kev , while @xmath31 describes a large interval 0.3 - 1.7 . the presence of the nonthermal component has the effect to slightly decrease the best fit value of the temperature ( @xmath32 kev ) , with respect to the temperature obtained considering only the mecs data . the flux of the nonthermal component is rather stable , @xmath33 in the 20 - 80 kev energy range , against variations of @xmath31 . the contribution of the nonthermal component to the thermal flux in the 2 - 10 kev energy range is @xmath34 for @xmath351.70 . the analysis of the two observations with effective exposure times of @xmath023 ksec ( february 1998 ) and @xmath048 ksec ( february 1999 ) for the pds does not show significant flux variations . these results and the fact that the two clusters with a detected hard x rays excess ( coma and a2256 ) both have radio halos , strongly support the diffuse nonthermal mechanism as responsible for the excess , as discussed in the next section . a2256 is the second cluster , after coma ( fusco - femiano 1999 ) , which shows hard x - ray radiation up to @xmath080 kev in the pds spectrum , with a clear excess above the thermal intracluster emission . ( a2199 shows only a marginal evidence in the external region of the mecs detectors , kaastra 1999 ) . we have investigated the possibility that the observed excess in a2256 could be due by a confusing source in the field of view of the pds . the most qualified candidate is the qso 4c + 79.16 observed by pspc with a count rate of @xmath00.041 c / s ( wga catalogue ) . with a typical photon index of 1.8 ( reports a steeper index of @xmath02.5 ) , about 1.2 c / s are necessary to account for the observed hxr emission of @xmath33 in the 20 - 80 kev energy range of the pds . considering that the qso is @xmath36 off - axis , an unusual variability of about two orders of magnitude is required . there is still the possibility that an obscured source , like circinus ( matt 1999 ) , be responsible of the detected hxr radiation . unless the obscured source is within 2@xmath21 of the central bright core of a2256 , our analysis of the mecs image excludes the presence of this kind of sources in the central region ( @xmath10 in radius ) of the cluster . the application of the inverse compton model , based on the scattering of relativistic electrons with the 3k background photons , appears less straightforward in a2256 than in the coma cluster . the radio morphology is remarkably complex ( bridle & fomalont 1976 ; bridle 1979 ; rottgering 1994 ) . there are at least four radio sources classified as head - tail radio galaxies , an ultra steep spectrum source and a diffuse region in the north with two diffuse arcs ( @xmath37,@xmath38 according to bridle 1979 ) , at a distance of @xmath39 from the cluster centre . the extent of this diffuse region is estimated to be [email protected] mpc , with a total flux density of 671 mjy at 610 mhz and a rather uniform spectral index of [email protected] between 610 and 1415 mhz ( bridle 1979 ) . the percentage polarization is uniform with an average value of 20% . the alignement of the electric field vectors suggests a well ordered magnetic field . the equipartition magnetic field is 1 - 2@xmath41 ( bridle 1979 ) . a fainter extended emission permeates the cluster centre ( diffuse emission around @xmath42 in bridle 1979 ) with a steeper radio spectral index of @xmath01.8 as estimated by bridle ( 1979 ) and in agreement with the 327 mhz data from the westerbork northern sky survey ( rengelink 1997 ) . the total flux density is 100 mjy at 610 mhz and no polarized emission have been detected from this region . we note that the physical and morphological properties of the diffuse @xmath42 emission are consistent with those of central halo sources while those in the @xmath43 region are consistent with the properties of peripheral relic sources as 1253 + 275 in the coma cluster . in addition to the thermal emission , a second component in the x - ray spectrum of a2256 was noted by markevitch & vikhlinin ( 1997 ) in their spectral analysis of the data in the central r=@xmath44 spherical bin . although they were not able to firmly establish the origin of this emission , their best fit is a power law model with photon index [email protected] , therefore favouring a nonthermal component . the contribution of this component to the total flux is not reported in the paper . considering that there are no bright point sources in the hri image , they argued for an extended source . also the joint gis & pca data analysis is consistent with the detection of a nonthermal component in addition to the thermal component . the contribution of this nonthermal component to the total x - ray flux in the 2 - 10 energy range is @xmath45 . however , a second thermal component ( 0.75 - 1.46 kev ) , instead of a nonthermal one , provides a better description of the data ( henriksen 1999 ) . the mecs data do not show evidence of this steep nonthermal component in the central bin of @xmath46 because the energy range is truncated to a lower limit of 2 kev ( molendi , de grandi , & fusco - femiano 2000 ) , while a joint fit to the lecs & mecs data within @xmath47 does not show a significant evidence for an additional component at energies lower than 2 kev . the power - law component ( slope @xmath48 ) found in the analysis of the data ( markevitch & vikhlinin 1997 ) , and the upper limit of 1.7 for @xmath49 , determined by data , suggest that two tails could be present in the x - ray spectrum of a2256 . the former might due to the central diffuse radio source with the steep index @xmath501.8 , and the last to the more extended radio emission in the northern region of the cluster with the flatter energy spectral index of [email protected] . assuming that the contribution of the power - law component , detected by , to the total x - ray flux ( @xmath51 ) is @xmath52 , we obtain a negligible contribution at pds energies ( @xmath53 ) and a magnetic field in the central radio region of @xmath3 . for the external radio region , with spectral index 0.8 , the nonthermal x - ray flux @xmath54 , derived by the pds excess , leads to a low value of @xmath2 . even assuming that a large fraction ( say 50@xmath55 ) of the hxr flux is due to the several point radio sources in the central region and/or to the contribution of different mechanisms , we obtain only a slightly greater value of @xmath56 . the combined fit of gis and pca data ( henriksen 1999 ) gives an upper limit of @xmath57 in the 2 - 10 kev energy range for the nonthermal component that corresponds to a lower limit for the volume - averaged intracluster magnetic field , @xmath58 , of 0.36@xmath41 ( @xmath59 ) . considering that the hxr flux detected by the pds is in agreement with the above value , we would obtain a value for @xmath58 consistent with that derived by the gis & pca data , but the fit to the mecs & pds data is unacceptable for @xmath60 . the previous scenario of a decreasing intracluster magnetic field from the cluster center would be difficult to reconcile with the stronger periferal radio region and higher equipartition magnetic field with respect to the central radio halo . therefore , we could consider the possibility , recently suggested by brunetti ( 1999 ) , that the hxr ic spectrum may be flatter than the synchrotron radio spectrum because of the acceleration and energy loss processes that produce an electron spectrum with different slopes . a different electron spectrum index for hxr and radio emissions is more likely for low magnetic fields which require higher electron energies for synchrotron than for ic radiation . this could explain the better fit to the pds data of a2256 with @xmath61=1.8 . besides , this model suggests an alternative interpretation of the hxr excess of a2256 . we can consider that a single hard tail is present in the x - ray spectrum of the cluster with index @xmath351.7 , as detected by the pds . the electron spectrum responsible of this hxr ic emission can produce radio emission with spectral index @xmath62 - 1=0.7 with a resulting mean volume - averaged intracluster magnetic field higher than the one we derive from the classical ic model . a different mechanism which may produce hxr radiation is given by nonthermal bremsstrahlung . sarazin & kempner ( 1999 ) suggest that all or part of the hxr emission detected in the coma cluster might be ntb from suprathermal electrons formed through current acceleration of the thermal gas , either by shocks or turbulence in the icm . for a2256 the mecs & pds measurements determine a power - law momentum spectrum of the electrons with index @xmath63 ( 90% ) . the consequence is that an accelerating electron model with flat spectrum produce more ic hxr emission than the ntb mechanism , unless the electron spectrum cuts - off or steepens at high energies . besides , these models produce more radio emission than observed if @xmath58 is @xmath64 .	after the positive detection by of hard x - ray radiation up to @xmath080 kev in the coma cluster spectrum , we present evidence for nonthermal emission from a2256 in excess of thermal emission at a 4.6@xmath1 confidence level . in addition to this power law component , a second nonthermal component already detected by could be present in the x - ray spectrum of the cluster , not surprisingly given the complex radio morphology of the cluster central region . the spectral index of the hard tail detected by the pds onboard is marginally consistent with that expected by the inverse compton model . a value of @xmath2 is derived for the intracluster magnetic field of the extended radio emission in the northern regions of the cluster , while a higher value of @xmath3 could be present in the central radio halo , likely related to the hard tail detected by .
bright submillimeter - selected galaxies ( smgs * ? ? ? * ) provide a powerful probe into the distant universe . thanks to the negative @xmath5-correction in the rayleigh - jeans tail of the dust thermal emission , flux limited submillimeter surveys with 850 @xmath2 m flux density @xmath18 mjy reach an almost uniform integrated infrared ( ir ) luminosity limit across a wide redshift range ( @xmath19 ) and yield a galaxy population mostly at redshifts between @xmath20 @xcite . with star formation rates ( sfrs ) of @xmath21 @xmath22 yr@xmath13 , the smgs are the most intense star - forming galaxies , despite their inevitably short - lived nature ( lifetime @xmath23 0.1 gyr ) . although such intense starburst systems are extremely rare in the local universe , smgs and the lyman break galaxies may contribute equally to the comoving sfr density at @xmath24 @xcite . in addition to their unique energetics , they also represent an important stage in massive galaxy formation . multiple lines of evidence suggest that smgs are likely the progenitors of massive elliptical galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , which apparently have formed bulk of their stars rapidly at an early epoch ( e.g. , * ? ? ? our understanding of this important high - redshift galaxy population are limited by the sensitivity and spatial resolution of current facilities . gravitational lensing offers an elegant solution by effectively lifting both limiting factors . also thanks to the negative @xmath5-correction , it is relatively straightforward to identify strongly lensed smgs in large area submillimeter surveys . @xcite and @xcite predict that extragalactic sources with 500 @xmath2 m flux density @xmath25 mjy are mostly strongly lensed or blended smgs , nearby late - type galaxies , and radio active galactic nuclei ( agns ) . as demonstrated by @xcite , objects in the last two categories can be easily removed using data at other wavelengths , leading to an extremely high success rate in identifying strongly lensed smgs with this technique ( see also @xcite ) . this simple flux selection has produced a few well - studied strongly lensed smgs ( lockman01 @xmath26 : @xcite ; id141 @xmath27 : @xcite ; hls j091828.6@xmath28514223 @xmath29 : @xcite ; and hatlas12@xmath100 @xmath30 , the subject of this paper ) , all of which were discovered by the _ space observatory @xcite . more complex selection processes have been proposed ( e.g. , * ? ? ? * ) , which would allow selecting hundreds of fainter lensed galaxies with _ the brightest of the lensed smgs might also be detected by the _ planck _ mission @xcite . with @xmath34 resolutions , such sources are probably blended with fainter sources even in the highest frequency / resolution channels of _ planck _ ( 545 and 857 ghz , or 550 and 350 @xmath2 m ) . the phase - i 130 deg@xmath4 of the _ herschel_-astrophysical terahertz large area survey ( h - atlas ; * ? ? ? * ) covers 28 _ planck _ sources in the _ planck _ early release compact source catalog ( ercsc ; * ? ? ? @xcite find that sixteen of them are high galactic latitude cirrus , ten are low - redshift galaxies , and one is resolved into two similarly bright nearby spirals ( ngc 3719 and 3720 ) . only one _ planck _ source is dominated by high - redshift galaxies : plckerc857 g270.59 + 58.52 ( @xmath31 jy , @xmath32 jy ) . with 18 , 25 , and 36 angular resolutions at 250 , 350 and 500 @xmath2 m , respectively , _ herschel _ detect 16 objects within a 4.23 radius of the _ planck _ position ( _ planck _ has a full - width - half - maximum [ fwhm ] resolution of 4.23 at 857ghz ) . there are 15 faint ( @xmath33 mjy ) sources surrounding an unusually bright source ( hatlas j114637.9@xmath1001132 , hereafter hatlas12@xmath100 ; @xmath34 mjy ) . taking into account the differences in the beam size and the filter transmission , the _ herschel _ sources account for only @xmath328% and @xmath324% of the _ planck _ flux densities at 545 and 857 ghz , respectively , suggesting that the _ planck _ measurements are boosted because of either positive noise spikes ( i.e. , eddington bias ; * ? ? ? * ) or blending with an over - density of sources that are below the confusion limit of _ herschel _ @xcite . the reader is referred to @xcite for a detailed _ planck_-_herschel _ comparison . hatlas12@xmath100 peaks at 350 @xmath2 m in flux density ( `` 350 @xmath2 m peaker '' ) , implying a high photometric redshift given typical dust temperatures . subsequent detections of multiple carbon - monoxide ( co ) lines from this unusually bright object determined a redshift of @xmath35 ( zspectrometer , carma , z - spec ; * ? ? ? * van der werf et al . in prep ; riechers et al . in prep ) . the high 500 @xmath2 m flux density ( @xmath36 mjy ) , in combination with the confirmed high redshift , makes hatlas12@xmath100 an excellent strong lens candidate . it is also the only strongly lensed smg candidate associated with a _ detection in the entire 130 deg@xmath4 h - atlas phase - i region . although the _ planck _ detection is partly due to spurious factors ( i.e. , eddington bias and/or blending ) , the confirmation of the lensed nature of the dominating source demonstrates that _ planck _ can efficiently identify the brightest lensed smgs once galactic cirrus and low - redshift galaxies are removed . in this paper we present a detailed multi - wavelength analysis of this _ planck_-associated smg . we describe our high - resolution keck adaptive optics imaging , submillimeter array ( sma ) and jansky very large array ( jvla ) interferometric observations , and the panchromatic photometry in [ sec : obs ] . we then perform a joint strong lens modeling at rest - frame 0.5 @xmath2 m , 200 @xmath2 m , and co(1@xmath60 ) in [ sec : model ] . in [ sec : sed ] we derive the intrinsic physical properties of the smg from its spectral energy distribution ( sed ) . we conclude by discussing the physical properties of hatlas12@xmath100 in the context of unlensed @xmath17 smgs ( [ sec : summary ] ) . throughout we adopt a @xmath37cdm cosmology with @xmath38 , @xmath39 and @xmath40 = 70 km s@xmath13 mpc@xmath13 . we obtained a 3,440-s @xmath41-band ( hereafter @xmath5 ) image on 2011 april 13 ( ut ) and a 2,100-s @xmath42-band image on 2011 june 30 ( ut ) with the keckii laser guide - star adaptive - optics system ( lgsao ; * ? ? ? * ) . an @xmath43 magnitude star 48 sw of hatlas12@xmath100 served as the tip - tilt reference star . the estimated strehl ratios at the source position are @xmath323% and 5% in @xmath5 and @xmath42-band , respectively . we used the nirc2 camera at 004 pixel@xmath13 scale for both filters ( 40 field ) , and dithered with 2@xmath13 steps . the atmospheric seeing at 0.5 @xmath2 m was @xmath30.4 and 0.5 during the @xmath5 and @xmath42-band imaging , respectively . we used our idl ( interactive data language ) programs to reduce the images . after dark subtraction and flat - fielding , sky background and object masks are updated iteratively . for each frame , after subtracting a scaled median sky , the residual background is removed with b - spline models . in the last iteration , we discard the three frames of the poorest image quality and correct the nirc2 geometric distortion using the solution of p. b. cameron before combining the aligned frames . the resolution of the final @xmath5 and @xmath42-band images are 016 and 027 in fwhm , respectively . we measure the fwhms from the most compact source in the field located 10 se of hatlas12@xmath100 ( labeled `` psf '' in fig . [ fig : obs]@xmath44 ) ; we also use this object as the psf in the lens modeling ( [ sec : amoeba ] ) . the images are flux calibrated against ukirt infrared sky survey ( ukidss ; * ? ? ? * ) and reach depths of @xmath5 = 25.6 and @xmath45 ab for a 5@xmath46 detection with 01 and 02 radius apertures , respectively . limited by the small field of nirc2 , a deep wide - field image is required for astrometry calibration . optical imaging was obtained with the high - throughput auxiliary - port camera ( acam ) mounted at a folded - cassegrain focus of the 4.2-m william herschel telescope @xcite on 2011 april 26 ( ut ) . we obtained four images of 200 s on a @xmath32 field centered on hatlas12@xmath100 , without any filter . the seeing was @xmath309 . the images were reduced and combined following standard techniques in iraf . no accurate photometric calibration is possible because we did not use any broad - band filter . but by comparing sources extracted from the acam image and the sdss @xmath47-band catalog in the same field , we find that our image reaches an equivalent @xmath47-band 5-@xmath46 depth of 24.6 ab , or 2.3 magnitudes deeper than the sdss . we solve the astrometry of the acam image using the on - sky positions of sloan digital sky survey ( sdss ; * ? ? ? * ) dr8 sources inside the field . we use the astrometry routines in marc buie s idl librarybuie / idl/ ] to correct for offsets , rotation , and distortions , with four terms ( constant , @xmath48 , @xmath49 , and @xmath50 ) . sources that appear blended in the sdss catalog are excluded . with 35 sdss sources , we measure 1@xmath46 dispersions of @xmath51ra = 013 and @xmath51dec = 014 between the astrometry calibrated acam image and the sdss . finally , we use the same routines to match the nirc2 images to the acam image with 13 well - detected sources inside the 40 nirc2 field of view . the corrected nirc2 images show 1@xmath46 dispersions of @xmath51ra = 004 and @xmath51dec = 005 . we obtained sma interferometric imaging of hatlas12@xmath100 at 880 @xmath2 m ( 339.58 ghz ) in the compact array configuration with an on - source integration time ( @xmath52 ) of 1 hr and at 890 @xmath2 m ( 336.9 ghz ) in the subcompact array configuration with @xmath52 = 2 hr . the compact and subcompact observations took place on 2011 may 2 and 2012 january 14 , respectively . during both nights , atmospheric opacity was low ( @xmath53 ) and phase stability was good . both observations used an intermediate frequency coverage of 48 ghz and provide a total of 8 ghz bandwidth ( considering both sidebands ) . the quasars 1229@xmath28020 and 1058@xmath28015 were used for time - variable gain ( amplitude and phase ) calibration . the blazar 3c 279 served as the primary bandpass calibrator . for the compact data , we used titan as the absolute flux calibrator . for the subcompact data , we intended to use callisto as the flux calibrator , but jupiter might have fallen into one of the side lobes of the sma primary beam while we observed callisto . so we decided to use 3c 279 in lieu of callisto as the flux calibrator . it is possible to use 3c 279 because we have reliable measurements of its flux both before and after the observation of hatlas12@xmath100 . we used the invert and clean tasks in the multichannel image reconstruction , image analysis , and display ( miriad ) software @xcite to invert the _ uv _ visibilities and deconvolve the dirty map , respectively . we used natural weighting to obtain the best sensitivity . for the compact data , the cleaned image has a synthesized beam with a fwhm resolution of @xmath54 at a position angle ( pa ) of @xmath55 degrees east of north ; for the subcompact data , the beam is @xmath56 at pa = 65.7 deg . the primary beam of the sma is @xmath337 . the r.m.s . noise levels are 3.0 mjy beam@xmath13 and 3.6 mjy beam@xmath13 for the compact image and the subcompact image , respectively . hatlas12@xmath100 is resolved into two components by the sma ( fig . [ fig : obs ] ) . taking into account the 10% flux calibration uncertainty , the total flux is @xmath57 mjy and @xmath58 mjy for the compact image and the subcompact image , respectively . the latter agrees well with the laboca bolometer array flux measurement at 870 @xmath2 m ( [ sec : photometry ] ) . the compact array data did not fully capture the source flux because of the sparser array configuration , i.e. , @xmath59 of the total flux is distributed on spatial scales larger than those accessible to the sma in its compact array configuration . so we use the total flux from subcompact data for sed modeling ( [ sec : sedsmg ] ) . we chose to use the compact image for lens modeling ( [ sec : sma ] ) because of its higher spatial resolution . we find that using the subcompact image or the subcompact@xmath28compact combined image does not change the lens modeling result , but they give larger errors for the derived parameters . we exploited the recent upgrade to the national radio astronomy observatory s very large array @xcite , which includes the provision of ka - band receivers ( 26.540ghz ) , to observe the redshifted co(1@xmath60 ) emission from hatlas12@xmath100 at 27.06532ghz ( @xmath60ghz ; @xcite ) . observations were carried out dynamically during excellent weather conditions on 2012 january 6 and 8 . during this open shared risk observing period the available bandwidth from the new wideband interferometric digial architecture ( widar ) correlator consisted of two independently tunable output pairs of eight sub - bands each , with @xmath61-mhz full - polarisation channels per sub - band , giving a total bandwidth of 2,048mhz . at the redshift of hatlas12@xmath100 , however , the co(1@xmath60 ) line could be reached by only the bd output pair , giving @xmath62 km s@xmath13 coverage and @xmath63 km s@xmath13 resolution . we offset our tuning by 64mhz to avoid noisier edge channels . the 8 sub - bands of output pair ac were tuned to 32.5ghz . the bright compact calibration source , j1150@xmath10023 were observed every few minutes to determine accurate complex gain solutions and bandpass corrections . 3c286 ( @xmath64 = 2.1666 jy at 27.06 ghz ) was also observed to set the absolute flux scale , and the pointing accuracy was checked locally every hour . in total , around 2hr of data were obtained for hatlas12@xmath100 , with @xmath31hr of calibration . the data were reduced using @xmath65 ( 31dec12 ) following the procedures described by @xcite , though with a number of important changes : data were loaded using bdf2aips and fring was used to optimize the delays , based on 1min of data for 3c286 . the base bands were knitted together using the noifs task , yielding @xmath66 datasets with 512 @xmath67 2-mhz channels , which we then added together using the task dbcon . finally , the channels were imaged over a @xmath68 field , with natural weighting ( robust = 5 ) , to form a @xmath69 cube centered on hatlas12@xmath100 . integrating over those 55 channels found to contain line emission ( so a fwzi of @xmath70 km s@xmath13 ) yielded an r.m.s . noise level of 27@xmath2jybeam@xmath13 . the cleaned and velocity - integrated co map is shown in fig . [ fig : obs]@xmath71 . the beam is 25@xmath6722 at pa = @xmath72 . similar to the sma , the map resolves two components separated by @xmath35 . the co lines extracted from the two components show the same redshift and line profile , further confirming that they are lensed images of a single source . the best - fit gaussian to the area - integrated spectrum gives a line width of @xmath73 km s@xmath13 and a line flux of @xmath74jykm s@xmath13 . in comparison , the co(1@xmath60 ) measurements reported by @xcite using zpectrometer on the green bank telescope are : @xmath75 km s@xmath13 and @xmath76jykm s@xmath13(corrected for the 20% difference in the absolute flux density of 3c 286 ) . the reason for the discrepancy is unclear , but the two line flux measurements agree within the 1@xmath46 errors . so hereafter , we use the weighted mean of the two measurements , @xmath77jykm s@xmath13 , to derive the molecular gas mass . llccc sdss & @xmath78 & 0.36 & @xmath79 & + & @xmath80 & 0.47 & @xmath81 & + & @xmath82 & 0.62 & @xmath83 & + & @xmath47 & 0.75 & @xmath84 & + & @xmath85 & 0.89 & @xmath86 & + ukidss & @xmath49 & 1.03 & @xmath87 & + & @xmath42 & 1.25 & @xmath88 & @xmath89 + & @xmath90 & 1.63 & @xmath91 & + & @xmath5 & 2.20 & @xmath92 & @xmath93 + wise & @xmath94 & 3.35 & @xmath95 & @xmath96 + & @xmath97 & 4.60 & @xmath98 & @xmath99 + & @xmath100 & 11.56 & @xmath101 & @xmath102 + & @xmath103 & 22.09 & @xmath104 & + pacs & green & 100 & & @xmath105 + & red & 160 & & @xmath106 + spire & blue & 250 & & @xmath107 + & green & 350 & & @xmath108 + & red & 500 & & @xmath109 + laboca & & 870 & & @xmath110 + sma & & 890 & & @xmath111 + mambo & & 1200 & & @xmath112 + carma & & 2792 & & @xmath113 + & & 3722 & & @xmath114 + vla & & 214000 & & @xmath115 + photometry of hatlas12@xmath100 were obtained from the sdss ( @xmath116 and @xmath85 ) , the ukidss ( @xmath117 and @xmath5 ) , the wide - field infrared survey explorer ( _ wise _ , 3.6 and 4.6 @xmath2 m ; * ? ? ? * ) , the _ herschel_/pacs ( 100 and 160 @xmath2 m ; program i d : ot1_rivison_1 ; * ? ? ? * ) , the _ herschel_/spire ( 250 , 350 , and 500 @xmath2 m ; * ? ? ? * ; * ? ? ? * ) , the large apex bolometer camera ( laboca , 870 @xmath2 m ; * ? ? ? * ) , the sma ( 880 @xmath2 m ) , the max - planck millimetre bolometer ( mambo , 1.2 mm ; * ? ? ? * ) , the combined array for research in millimeter - wave astronomy ( carma , 2792 and 3722 @xmath2 m ; * ? ? ? * ) , and the vla first survey ( 21 cm ; * ? ? ? we obtained imaging at 870 @xmath2 m with the laboca bolometer array at the atacama pathfinder experiment ( apex ) telescope in november 2011 ( clements et al . in prep ) . laboca observed a 11.4 diameter field with a resolution of fwhm = 186 . the observations have a total integration time of @xmath330 hr reaching a 1@xmath46 sensitivity of @xmath32 mjy . we obtained 1.2 mm imaging with mambo at the iram 30-m telescope ( fwhm @xmath3 107 ) in january and february 2011 ( dannerbauer et al . in prep . ) . observing time in the on - off mode is 24 minutes , achieving a 1@xmath46 sensitivity of @xmath31 mjy . we obtained continuum observations at 81.2 and 108.2 ghz ( 3722 and 2792 @xmath2 m ; covering rest - frame co[3@xmath62 ] and co[4@xmath63 ] lines ) on 2011 mar 18 and sep 1 as part of our co follow - up campaign of bright , lensed h - atlas smgs with carma in d array ( riechers et al . in prep ) . observations were carried out for 0.9 and 1.4 hr on source , respectively , using the 3 mm receivers and a bandwidth of 3.7 ghz per sideband . hatlas12@xmath100 is unresolved in these observations , with angular resolutions of @xmath118 and @xmath119 at 81.2 and 108.2 ghz , respectively ( restored with natural baseline weighting ) . table [ tab : photo ] lists the photometry . we have included in the errors the absolute flux calibration uncertainties ( 3% for _ wise _ , 3@xmath15% for pacs , 7% for spire , 10% for sma , and 15% for laboca , mambo , and carma ) . because the lgsao image has the highest spatial resolution , we use it to find the best - fit lens model . we initially use the peak positions of the multiply - imaged source to constrain the lensing potentials ( [ sec : peak ] ) , then we exploit the @xmath5-band light distribution in the image plane to quantify the morphologies of the source as well as refining the lensing potentials ( [ sec : amoeba ] ) . finally , we use the best - fit lensing potentials and the sma and jvla images to constrain the sizes and locations of the dust and molecular gas in the source plane ( [ sec : sma ] & [ sec : evla ] ) . we use lenstool @xcite to find the best - fit parameters and their errors from the peak positions . lenstool implements a bayesian markov chain monte - carlo sampler to derive the posterior distribution of each parameter and an estimate of the evidence for the model . the lensing system is mainly made of two red filaments that are @xmath335 apart ( fig . [ fig : obs]@xmath44 ) . the outward curved shape of the northern arc can be explained if the source is intrinsically curved . hence we split each of the two arcs into three parts and build a simple lens model by putting two deflectors centered on g1 and g2 . we find that the predicted counter - images can explain the additional features close to g1 and g2 . guided by the predicted counter - images , we define three systems of lensed images ( fig . [ fig : obs]@xmath44 @xmath120 ) . the 11 peak positions in 3 separate systems provide a total of 16 constraints ( @xmath121 ) , allowing us to include shear from nearby galaxies g3 and g4 . for the lensing galaxies , we find photometric redshifts of @xmath122 and @xmath123 with the public photo-@xmath85 code eazy @xcite . we obtain the nine - band photometry from the sdss ( @xmath116 and @xmath85 ) and the ukidss ( @xmath117 and @xmath5 ) surveys . at these wavelengths , the flux from the lensed galaxy is negligible ( [ sec : sed ] ) . the redshift of g1@xmath28g2 is measured from the total fluxes of g1 and g2 , because they are blended in the seeing limited data . g3 is undetected in sdss but shows similar color as g1 and g2 . hence in the lens modeling we assume all four galaxies are at @xmath124 . note that although redshift errors of the lensing galaxies would lead to errors in the estimated lens masses , they would not change our conclusions on the lensed galaxy because the magnification factors would remain the same . for the lens model , we assume that the dark - matter plus baryonic mass profiles of the foreground lens galaxies g1 to g4 can be described as singular isothermal ellipsoids ( sies ; * ? ? ? the sie profile is parametrized by the velocity dispersion ( @xmath46 ) , the position ( @xmath125 , @xmath126 ) , the axis ratio ( @xmath127 ) , and the pa ( @xmath128 , e of n ) . we fix the positions to the centers of the galaxies . for g3 and g4 , we further fix their @xmath129 and @xmath128 to those from the light distribution , because they are not well constrained by the peak positions and there are significant correlations between the pa and ellipticity of the light and of the mass distribution ( e.g. * ? ? ? therefore , we have a total of eight free parameters . we find a best - fit with @xmath130 for dof = 8 ( degrees of freedom ) and an average positional error of 004 ( @xmath31 pixel ) . the parameters and their errors are summarized in table [ tab : lens ] . we also list the mass enclosed by the critical curve for each sie , @xmath131 where @xmath132 , @xmath133 , and @xmath134 are the angular diameter distances to the lens , to the source , and between the lens and the source , respectively . the radius of the area enclosed by the critical curve can be approximated by the circularized einstein radius : @xmath135 in the errors of masses and velocity dispersions , we have included the 1-@xmath46 uncertainty of the photometric redshift . the nominal model described above is the most favorable description of the lensing system because of the following : 1 . adding @xmath129 s and @xmath128 s of g3 and g4 as free parameters does not substantially improve the fit : the bayesian evidence , strong if @xmath136 , and decisive if @xmath137 @xcite . ] increases by only @xmath138 = 0.5 and the reduced @xmath139 actually increases from 1.0 to 1.6 as a result of the decreased degree of freedom . 2 . excluding the potentials of g3 and/or g4 does degrade the fit significantly . the bayesian evidence decreases by @xmath138 = 2.5 and 30 , and the reduced @xmath139 increases from 1.0 to 1.8 and 6.5 , when we exclude g4 and both g3 and g4 , respectively . including a group - scale potential with a piemd profile ( pseudo - isothermal elliptic mass distribution ; * ? ? ? * ) does not improve the fit . for the piemd profile , we adopt a cut - off radius of 500 kpc but allow the position , ellipticity , pa , core radius , and velocity dispersion to vary . to limit the number of free parameters , we fix the @xmath129 s and pa s of the sies to those measured from the lensing galaxies but allow their velocity dispersions to vary . so we have a total of 10 free parameters . the reduced @xmath139 of the best - fit is 1.9 , much higher than that of the nominal model . the bayesian evidence also decreases by @xmath138 = 6.9 when compared with the nominal model . llccr g1 & @xmath140 & @xmath141 & @xmath142 & + & @xmath143 & @xmath144 & @xmath145 & deg + & @xmath146 & @xmath147 & @xmath148 & km s@xmath13 + & @xmath149 & @xmath150 & @xmath151 & @xmath12 + g2&@xmath140 & @xmath152 & @xmath153 & + & @xmath143 & @xmath154 & @xmath155 & deg + & @xmath146 & @xmath156 & @xmath157 & km s@xmath13 + & @xmath149 & @xmath158 & @xmath158 & @xmath12 + g3&@xmath146 & @xmath159 & @xmath160 & km s@xmath13 + & @xmath149 & @xmath161 & @xmath162 & @xmath12 + g4&@xmath146 & @xmath163 & @xmath164 & km s@xmath13 + & @xmath149 & @xmath165 & @xmath166 & @xmath12 + & & & & + smg & @xmath167 & & @xmath168 & + & @xmath169 & & @xmath170 & + & @xmath171 & & @xmath172 & + although the peak positions can constrain the deflectors through ray tracing , they can not provide an accurate estimate of the magnification factor because the source - plane light distribution is not taken into account . because we want to estimate the intrinsic properties of the lensed galaxy , we are interested in the luminosity weighted magnification factor , which depends on the source morphology because the magnification factor is different at each source plane position . in this section we model the morphology of the lensed galaxy and refining the lensing potentials simultaneously with the @xmath5-band image . the psf is derived from the most compact source in the field , which is 10 se of hatlas12@xmath100 ( fig . [ fig : obs]@xmath44 ) . following [ sec : peak ] , we assume that the source consists of three clumps , each described as a srsic profile . again we use sie profiles for the lensing potentials . so we have a total of 29 parameters : seven parameters for each srsic profile , and eight parameters for the sie potentials . our fitting procedure is as follows . for an initial set of parameters describing the source and the lenses from [ sec : peak ] , we use lenstool to generate lensed images of the source , which is then convolved with the psf and compared with the observed image . we limit the comparison in a @xmath173 ( @xmath174 pixels ) rectangular region that encloses the lensing features . this process is iterated with amoeba_sa to find the parameters that minimize the residual between the observation and the model . amoeba_sa is based on the idl multidimensional minimization routine amoeba @xcite with simulated annealing added by e. rosolowsky . we allow a maximum of 1000 iterations in each call of amoeba_sa . for the simulated annealing , we adopt an initial `` temperature '' of 100 and decrease it by 20% in each subsequent call to amoeba_sa . a good fit with a reduced @xmath139 around unity is normally found after a few calls to amoeba_sa . for each iteration , we compute the total luminosity - weighted magnification factor ( @xmath175 ) by summing the pixel values in the image and the source planes with apertures matched by inverting the image plane aperture to the source plane . the 1-@xmath46 confidence interval of @xmath175 is found with @xmath176 . note that we compute the @xmath139 values on the residual image binned by @xmath177-pixel boxes ( fwhm = 016 = 4 pixel ) , so that the noise becomes uncorrelated between pixels ; or equivalently , one could divide the @xmath139 values from the original residual images by a factor of 16 . we find the luminosity - weighted magnification to be @xmath178 @xmath168 . the best - fit parameters for the deflectors are listed in table [ tab : lens ] . the results are very similar to those from fitting the peak positions , although the errors are smaller because the entire image provides more information than the peak positions alone . figure [ fig : lens ] shows the best - fit model . the lensed galaxy has a curved morphology , causing the northern arc bending in the opposite direction of the deflectors . in the source plane , the three clumps extend over only 021 or 1.6 kpc , and their effective radii are @xmath179 ( @xmath180 kpc ) , @xmath181 ( @xmath182 kpc ) , and @xmath183 ( @xmath184 kpc ) from w to e. the nature of the feature nne of g2 in the residual image is unclear , but it is unlikely to be at the same redshift as the lensed galaxy : tracing its position to the source plane and imaging it back predicts an _ unobserved _ equally bright counter - image 08 s of the southern arc . this feature could therefore be part of the galaxy g2 . precise astrometry calibration is crucial for a joint analysis of images from different wavelengths . because the only @xmath5 source detected by the sma is hatlas12@xmath100 , we have to estimate the astrometry offset between the two images in a statistical way . because the keck image is tied to the sdss astrometry and the sma image is tied to the radio reference frame , we cross - correlate the vla first catalog @xcite and the sdss catalog within @xmath185 of hatlas12@xmath100 and compute the optical - radio separation . ninety - four radio sources have optical counterparts within 3 . we then fit an elliptical gaussian to the two - dimensional distribution in @xmath186 and @xmath187 . the systematic offset from the peak position of the gaussian is consistent with zero ( @xmath188 , @xmath189 ) . the best - fit gaussian has @xmath46 s of 040 and 030 , and a pa of @xmath190 for the major axis . therefore , the 1@xmath46 ellipse of the astrometry offset has major / minor semi - axes of 061/045 . our result is consistent with that of @xcite , who found a @xmath301 systematic offset and a 1@xmath46 error circle of 047 in radius between first and sdss astrometry . we can constrain the astrometry offset further through lens modeling . as demonstrated by @xcite and @xcite , interferometric data are most naturally modeled with the @xmath66-plane visibilities , because it avoids beam deconvolution and naturally handles correlated noise . here , however , we opt to model the cleaned map directly , because ( 1 ) the images are essentially unresolved in the sma map , and ( 2 ) we already have a good lens model from the @xmath5-band image ( [ sec : amoeba ] ) . because of the limited spatial resolution of the sma 880 @xmath2 m image , the two centroid positions do not offer enough information to constrain the lens model . hence , for the deflectors , we fix all the parameters to the best - fit values from [ sec : amoeba ] ; for the source , we assume a circular gaussian profile with variable position and size . we shift the sma image relative to the @xmath5 model on a 2@xmath672 grid with 01 steps . at each offset position , we find the best - fit model using the same fitting procedure as in [ sec : amoeba ] . the modeling is performed on a @xmath191-pixel ( @xmath192 ) region enclosing the sma sources . figure [ fig : xychi2]@xmath44 shows a map of the minimum @xmath139 values at each offset position . the global best - fit , with reduced @xmath139 of unity , is reached when we shift the sma image @xmath193 e of the @xmath5 image . the middle panels of fig . [ fig : lens ] show this global best - fit model . the noise of the sma map is gaussian but is highly correlated . we compute the r.m.s . noise of the sma map after binning it by boxes of @xmath194 pixels . we find that the noise starts to decrease as @xmath195 for @xmath196 pixels ( fwhm @xmath197 2 = 20 pixel ) , indicating that the noise becomes uncorrelated on 20-pixel scales . therefore , we divide the @xmath139 values from the the residual images by a factor of 400 , which is equivalent to computing @xmath139 from residual images binned by @xmath198-pixel boxes . in combination with the 1@xmath46 error ellipse from first sdss cross - correlation , we determine that the astrometry offset between 880 @xmath2 m and @xmath5 images is @xmath199 and @xmath200 ; i.e. , the overlapping region between the ellipse and the 1@xmath46 contour of the @xmath139 map . collecting all of the solutions in this permitted offset region satisfying @xmath176 , we estimate a luminosity - weighted 880 @xmath2 m magnification of @xmath201 @xmath170 , and an 880 @xmath2 m source size of fwhm = @xmath202 = @xmath203 kpc . because we have fixed the deflectors with the best - fit parameters from @xmath5-band , the errors here do not include the uncertainties of the deflectors . higher resolution far - ir images are required to constrain the deflectors and the source simultaneously . dust emitting regions are often spatially offset from the uv / optical emitting regions in smgs @xcite . this is clearly the case for hatlas12@xmath100 , which shows distinctly different morphologies at @xmath5-band and 880 @xmath2 m , even after convolving the @xmath5-band image with the sma beam ( compare fig . [ fig : lens]@xmath71 & @xmath204 ) . from the lens model , we estimate a source - plane separation between the 880 @xmath2 m source and the central @xmath5 clump of @xmath205 or @xmath206 kpc ( fig . [ fig : lens]@xmath207 ) . if we assume zero astrometry offset between sma and keck , then we obtain a model that poorly fits the observation ( @xmath208 ; fig . [ fig : xychi2]@xmath44 ) . the lens model gives a slightly larger magnification ( @xmath209 ) and doubles the source size ( fwhm = @xmath210 kpc ) . however , the source - plane separation between the 880 @xmath2 m source and the central @xmath5 clump remains the same ( @xmath211 kpc ) . we use the same technique to model the jvla co(1@xmath60 ) map as in [ sec : sma ] . the lensed images are better resolved in the jvla image than in the sma image , so we use an elliptical gaussian instead of a circular gaussian for the source profile . the model has a total of six free parameters ( @xmath212 , fwhm , @xmath129 , pa , and flux density ) . again , we can constrain the astrometric offset between jvla and keck through lens modeling . figure [ fig : xychi2]@xmath213 shows the minimum @xmath139 values at each offset position relative to the @xmath5 image . to deal with the correlated noise , we scale the @xmath139 values from the residual map by the product of the fwhms of the major and minor axes of the beam . the global best - fit , with reduced @xmath139 of unity , is reached when we shift the jvla image 02 e of the @xmath5 image . the bottom panels of fig . [ fig : lens ] show the best - fit model . we estimate a co magnification of @xmath214 @xmath172 . similar to the sma image , we find a source - plane separation of @xmath215 kpc between the cold molecular gas and the stellar emission ( i.e. , the central @xmath5 clump ) . the co(1@xmath60 ) is emitted from a more extended region than the dust , but the two spatially overlap ( fig . [ fig : lens]@xmath207 ) . the co source has fwhm = @xmath216 = @xmath217 kpc along the major axis , with an axis ratio of @xmath218 . the molecular gas disk is massive . the velocity - area - integrated co brightness temperature of @xmath219\times10^{11}$]k km s@xmath13 pc@xmath4 indicates a molecular gas reservoir of @xmath220\times10^{10}$]@xmath12 after lensing correction , assuming a conversion factor of @xmath221 @xmath12 ( k km s@xmath13 pc@xmath4)@xmath13 , which is commonly assumed for starburst environments @xcite . note , however , that @xmath222 is uncertain by at least a factor of a few and it may depend on the metallicity , the gas temperature , and the velocity dispersion of the galaxy @xcite . the co magnification factor determined from the lens model ( @xmath172 ) is in excellent agreement with that estimated from the co luminosity@xmath1fwhm correlation . the observed @xmath223 and line width indicates a magnification factor of @xmath224 , based on its deviation from the correlation established by unlensed smgs ( * ? ? ? * bothwell et al . in prep ) . this agreement demonstrates that strongly lensed smgs may be effectively selected with co spectroscopy in the future . llcr g1@xmath28g2 & @xmath225 & @xmath226 & @xmath12 + & age & @xmath227 & gyr + & @xmath228 & @xmath229 & gyr + & @xmath230 & @xmath231 & mag + & sfr@xmath232 & @xmath233 & @xmath234 + & & & + smg@xmath235 & @xmath225 & @xmath236\times10^{10}$ ] & @xmath12 + & sfr@xmath232 & @xmath237 & @xmath234 + & & & + smg@xmath238 & @xmath239 & @xmath240 & k + & @xmath241 & @xmath242 & + & @xmath243 & @xmath244\times10 ^ 8 $ ] & @xmath12 + & @xmath245 & @xmath246 & @xmath2 m + & @xmath46 & @xmath247 & kpc@xmath4 + & @xmath248 & @xmath249\times10^{13}$ ] & @xmath10 + & sfr@xmath250 & @xmath251 & @xmath234 + & & & + smg@xmath252 & @xmath239 & @xmath253 & k + & @xmath241 & @xmath254 & + & @xmath243 & @xmath255\times10 ^ 9 $ ] & @xmath12 + & & & + smg@xmath256 & @xmath223 & @xmath257\times10^{10}$ ] & k km s@xmath13 pc@xmath4 + & @xmath258 & @xmath259 & @xmath10/k km s@xmath13 pc@xmath4 + & fwhm@xmath260 & @xmath261 & km s@xmath13 + & @xmath262 & @xmath263\times10^{10}$ ] & @xmath12 + & @xmath264 & @xmath265\times10^{11}$ ] & @xmath12 + & @xmath266 & @xmath267% & + & @xmath268 & @xmath269% & + useful physical parameters are encoded in the seds . the optical - to - nir sed of hatlas12@xmath100 is dominated by the foreground galaxies g1 and g2 , from which we can derive the photometric redshift and the stellar population of the lensing galaxies . the far - ir and submillimeter regime is dominated by the lensed smg , as evident in the sma image , so the data can tell us the dust and star formation properties of the smg . adopting the photometric redshift of 1.06 , we model the nine - band photometry ( @xmath78 , @xmath80 , @xmath82 , @xmath47 , @xmath85 , @xmath49 , @xmath42 , @xmath90 , and @xmath5 ) of g1+g2 with the stellar population synthesis models of ( * ? ? ? we assume a @xcite initial mass function ( imf ) , @xcite extinction law , and exponentially declining star formation history , with a range of e - folding times ( @xmath228 = 0.1 to 30 gyr ) and ages ( 0.01 to 12.5 gyr ) . for each template , we fit for the stellar mass ( @xmath225 ) and extinction ( @xmath230 ) . the best - fit model gives @xmath270 for dof = 7 ( fig . [ fig : sed]@xmath44 ) . the derived properties of g1@xmath28g2 are listed in table [ tab : sedfit ] . the intrinsic extinction is small ( @xmath271 ) and there is very little current star formation ( sfr = @xmath233 @xmath234 ) . the dust - absorbed uv / optical luminosity ( @xmath272\times10^{10}$ ] @xmath10 ) is less than 0.15% of the total integrated ir luminosity before lensing correction ( @xmath273 = @xmath274 @xmath10 ) . therefore , g1 and g2 do not contribute significantly to the far - ir fluxes , in agreement with their absence in the sma image . the stellar mass from sed modeling is @xmath380% of the total mass within the critical curves from lens modeling ( table [ tab : sedfit ] ) , implying that the galaxies are dominated by stellar mass within @xmath275 kpc . in the @xmath42 and @xmath5-bands , we obtain the photometry of the smg with an aperture contoured around the multiply - imaged features after subtracting the foreground lenses . we measure @xmath276 and @xmath277 in ab magnitudes , consistent with the red @xmath278 colors of unlensed smgs @xcite . careful modeling is required to extract smg photometry from the _ wise _ data , because the smg is blended with the foreground galaxies g1 through g4 ( fwhm = 6@xmath112 ) . we model the _ wise _ 3.4 @xmath2 m source with four elliptical gaussians of the same shape . with galfit @xcite , we fix their positions to those determined from the keck image , but we allow the gaussian shape to vary . then we measure the flux density of g1@xmath28g2 ( @xmath3205 @xmath2jy ) decomposed from that of g3 and g4 ( @xmath3117 @xmath2jy ) . finally , the flux density of the smg ( @xmath337 @xmath2jy ) is estimated from the excess of g1@xmath28g2 relative to their best - fit stellar population synthesis model ( [ sec : sedlens ] ) . unfortunately , we can not separate g3@xmath28g4 from g1@xmath28g2 in the longer wavelength channels of _ wise _ because of the inferior image quality , so we treat the excesses over the best - fit model of g1@xmath28g2 as upper limits for the smg . the far - ir - to - millimeter sed is dominated by the lensed smg , therefore no foreground subtraction is necessary . table [ tab : photo ] summarizes the photometry for hatlas12@xmath100 . we opt to model the rest - frame optical and far - ir emission separately , instead of fitting them together in a self - consistent way with magphys @xcite , because our lens model shows that they are emitted from physically distinctive regions , i.e. , the dust that attenuates the optical emission has little to do with the starburst - heated dust that emits in the far - ir . it is difficult to constrain the stellar population with only three photometric detections ( dof = 1 ) in rest - frame optical . however , we can limit the parameter space by excluding unphysical models , such as those that require negative extinction corrections and those whose ages exceed the cosmic age at @xmath279 ( 1.9 gyr ) . we use the same bc03 templates as in the previous section . the shaded region in fig . [ fig : sed]@xmath44 shows all of the permitted models with @xmath280 . these models give a range of extinctions , stellar masses , and sfrs : @xmath281 , @xmath282\times10^{10}$ ] @xmath12 , and sfr @xmath283 @xmath234 ( table [ tab : sedfit ] ) . the dust - absorbed uv / optical luminosity range from 0% to 140% of the observed @xmath273 ; but 90% of the models have dust - absorbed luminosity less than 50% of the observed @xmath273 . we fit the far - ir sed with a single - temperature modified blackbody , @xmath284 where @xmath46 is the total absorption cross section of dust particles at the optically thick limit ( i.e. , the size of the dust - obscured region ) , @xmath285 the _ planck _ function , @xmath286 the optical depth , @xmath2 the lensing magnification factor , and @xmath287 the luminosity distance . in the optically thin limit ( @xmath288 ) , dust mass can be derived based on the knowledge of the opacity @xmath289 ( absorption cross section per unit mass ) : @xmath290 it is generally assumed that the opacity follows a power law , @xmath291 , and has a normalization of @xmath292 m@xmath4 kg@xmath13 at 850 @xmath2 m @xcite . both the general `` optically thick '' ( @xmath293 ) and the optically thin ( @xmath294 ) models provide good fits to the observed sed ( fig . [ fig : sed]@xmath213 ) . for the general model , we use all of the nine detections between 100 @xmath2 m and 3 mm . the best - fit general model gives @xmath295 for dof = 5 , suggesting that the photometric errors have been overestimated . for the optically thin model , we exclude the pacs 100 @xmath2 m point , which is clearly on the wien tail where small grains tend to dominate the emission . the best - fit optically thin model gives @xmath296 for dof = 5 . the derived parameters are listed in table [ tab : sedfit ] . the optically thick model yields dust properties similar to those of the local ultra - luminous infrared galaxy ( ulirg ) arp 220 @xcite , with the optical depth exceeds unity below rest - frame @xmath3250 @xmath2 m . the intrinsic @xmath297 @xmath2 m luminosity of @xmath298\times10^{13}$ ] @xmath10 classifies hatlas12@xmath100 as a hyper - luminous infrared galaxy ( hy - lirg ) . the ir luminosity implies an sfr of @xmath251 @xmath12 yr@xmath13 for a @xcite imf @xcite . using the values of @xmath273 and @xmath299 k in the stefan - boltzmann law we obtain a spherical source radius of @xmath300 pc , which is three times larger than that of arp 220 ( 230 pc ) because of the ten times greater luminosity . the source radius is comparable to that we derive from the optically thick model ( @xmath301 pc ) and is consistent with the size we measure from modeling the sma image ( fwhm = @xmath203 kpc ; [ sec : sma ] ) . therefore , the optically thick model is preferred . the radio luminosity from the observed 1.4 ghz flux density is @xmath302\times10^{25}$ ] w hz@xmath13 for a radio spectral index of @xmath303 . assuming the radio emission is magnified by the same factor as the submillimeter emission , the ir - to - radio luminosity ratio of hatlas12@xmath100 , @xmath304 , is consistent with the radio far - ir correlation of high - redshift starburst galaxies : e.g. , @xcite measured @xmath305 for 15 smgs , while @xcite measured @xmath306 for 65 _ herschel _ 250 @xmath2 m selected galaxies . this suggests that the agn contribution is insignificant in hatlas12@xmath100 . we also do not see significant agn contribution in the mid - ir . in fig . [ fig : sed]@xmath213 , we fit the seds of the local ulirgs mrk 231 , arp 220 , and the @xmath307 smg `` cosmic eyelash '' @xcite to the far - ir sed . the _ wise _ upper limits lie well below the agn - dominated ulirg mrk 231 but are more consistent with arp 220 and the eyelash . therefore , we conclude that hatlas12@xmath100 is predominantly a starburst system . we have presented high - resolution @xmath5-band , 880 @xmath2 m , and co(1@xmath60 ) observations and the near - ir - to - centimeter sed of a _ herschel_-selected strongly lensed smg at @xmath279 ( hatlas12@xmath100 ) . the smg shows distinctly different morphologies in the three images , suggesting differential magnification due to stratified morphologies . a joint strong lens modeling shows that the smg is lensed by four galaxies at @xmath7 and the luminosity - weighted magnification factors are @xmath168 in @xmath5 , @xmath170 at 880 @xmath2 m , and @xmath172 at co(1@xmath60 ) . in the source plane , the smg consists of several stellar clumps extended over @xmath31.6 kpc with @xmath236\times10^{10}$ ] @xmath12 of stars , a compact ( @xmath31 kpc ) starburst enshrouded by @xmath244\times10 ^ 8 $ ] of dust at @xmath360k , and an extended ( @xmath36 kpc ) cold molecular gas reservoir with @xmath263\times10^{10}$ ] @xmath12 of gas . the starburst and its gas reservoir are located @xmath34 kpc from the stars . similar separations between optical and submillimeter / radio emission have been observed in unlensed high - redshift dusty starbursts ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? . however , these previous results could also be attributed to astrometry offsets across different facilities ; @xcite might be the only exception , whose tied the astrometry of the images to larger radio and optical fields . in hatlas12@xmath100 , the physical separations amongst stars , dust , and gas are less ambiguous because of the clear wavelength - dependent morphologies in the image plane : fortuitously , the stars and dust / gas straddle across the caustic , so the less obscured @xmath5-band region is quadruply imaged while the heavily obscured starburst and its gas reservoir are doubly imaged . because of the @xmath34 kpc separation between the stars and the gas - rich starburst and their similar masses , it is tempting to suggest that the smg is in the process of a major merger , which presumably is driving the starburst activity in @xmath17 smgs ( e.g. , * ? ? ? * ; * ? ? ? however , spatial separation of this scale can also be explained by differential dust obscuration in a single galaxy , as has been proposed for other high - redshift smgs where spatial offsets have been observed between rest - frame uv and submillimeter ( e.g. , gn20 and aztec 3 ; * ? ? ? * ; * ? ? ? differential magnification may affect the observed far - ir sed as well as the co ladder , as hotter dust and higher-@xmath42 co lines may have more compact morphologies @xcite . however , because the magnification map is smooth in the area , the six times difference in the sizes of the co(1@xmath60 ) and the dust emitting region only lead to a @xmath310% difference in magnification , which is smaller than the 1@xmath46 errors of our estimates of the magnification factors . therefore , differential magnification is unlikely to be significant enough to affect the far - ir sed and future co ladder measurements in hatlas12@xmath100 . how does the massive gaseous disk compare with the disks in other smgs ? we estimate an dynamical mass of @xmath265\times10^{11}$]@xmath12 and a gas fraction of @xmath308% for the co(1@xmath60 ) disk using the `` isotropic virial estimator '' ( e.g. , * ? ? ? * ) : @xmath309 where @xmath310 is the co line fwhm in km s@xmath13 and @xmath311 is half of the fwhm size of the disk in kpc . combined with the fwhm disk radius , we further estimate a gas surface density of @xmath312 @xmath12 pc@xmath313 . both @xmath314 and @xmath315 are similar to those of the extended co(1@xmath60 ) disks in the two @xmath316 smgs in @xcite . but both values are significantly smaller than those of the kinematically resolved co(6@xmath65 ) disk of the `` cosmic eyelash '' ( @xmath317 , @xmath318@xmath12pc@xmath313 ; * ? ? ? * ) . the discrepancies illustrate the limitations of these widely used but crude estimators _ and/or _ that high excitation co lines probe more compact and denser regions in a disk . higher resolution observations are clearly needed to resolve this issue . without spatially resolved gas kinematics , we refrain from estimating the disk stability parameter of @xcite . hatlas12@xmath100 is a gas - rich , initial starburst system similar to unlensed smgs and local ulirgs . its intrinsic ir luminosity well exceeds @xmath319 @xmath10 ( i.e. , hy - lirg ) , implying an enormous rate of star formation ( @xmath251 @xmath12 yr@xmath13 ) . although the molecular gas reservoir is massive and it constitutes @xmath267% of the _ visible _ baryonic mass ( @xmath320 ) and @xmath269% of the dynamical mass , it will exhaust in just @xmath321 myr at the current sfr ( assuming no gas accretion ) . the star formation timescale , @xmath322/sfr = @xmath323 myr , is only @xmath31% of the cosmic age at @xmath279 ( @xmath324 gyr ) , suggesting that hatlas12@xmath100 is an initial starburst system with maturity @xmath325 @xcite . the specific sfr , ssfr = sfr/@xmath225 = @xmath326 gyr@xmath13 , is consistent with the average @xmath17 smgs , but it is an order of magnitude higher than the median value of the star - forming main sequence of lyman break galaxies at the same epoch @xcite . we can also estimate the star formation efficiency : @xmath327 , where @xmath328 is the dynamical or free - fall timescale . for @xmath329 = 3 kpc and @xmath330 , we obtain @xmath331 , or [ 11@xmath3324]% per dynamical timescale , which is comparable to unlensed smgs but is an order of magnitude higher than normal starforming galaxies @xcite . the dust mass of @xmath244\times10^{8}$ ] @xmath12 is similar to the average dust mass of unlensed smgs ( e.g. , * ? ? ? * ) , and the gas - to - dust ratio , @xmath333/0.07 m@xmath4 kg@xmath334/0.8 ) , is comparable to that of the milky way . assuming that the dust emission is indicative of the size of the starburst and a starburst disk radius of @xmath335 kpc , the star formation surface density of @xmath336 @xmath234 kpc@xmath313 approaches the eddington limit of radiation pressure supported starburst disks ( @xmath337 @xmath234 kpc@xmath313 ; * * ; * ? ? ? * ) , similar to local ulirgs such as arp 220 and the host galaxy of the @xmath338 quasar sdss j114816.64 + 525150.3 @xcite . in conclusion , hatlas12@xmath100 is a _ bona fide _ smg with an intrinsic submillimeter flux density of @xmath339 mjy . the starburst disk , where most of the molecular gas and dust reside , is spatially separated from the less obscured stellar population by @xmath34 kpc , suggesting either a major merger or differentiated dust obscuration . the @xmath31 kpc radius starburst disk is presumably supported in large by radiation pressure on the dust grains . its physical properties , such as molecular gas mass , stellar mass , gas - to - dust ratio , gas fraction , sfr , star formation efficiency , and radio - to - far - ir luminosity ratio , are all very similar to unlensed @xmath17 smgs @xcite . the lensing boost of the effective angular resolution and sensitivity has allowed us to examine in unprecedented details the properties of a typical starburst galaxy when the universe is only 1/7 of its current age . hatlas12@xmath100 provides a prelude to a golden age of smg research , as _ herschel _ is unveiling hundreds of strongly lensed smgs before the mission completes ( e.g. , * ? ? ? * wardlow et al . in prep ) . we thank the anonymous referee for comments that helped improve the paper . hf , ac , jlw and sk acknowledge support from nsf career ast-0645427 . ipf is supported by the spanish grants esp2007 - 65812-c02 - 02 and aya2010 - 21697-c05 - 04 . sgd acknowledges a partial support from the nsf grant ast-0909182 . gdz and mn acknowledge support from asi / inaf agreement i/072/09/0 ( `` _ _ planck _ _ lfi activity of phase e2 '' ) and from miur through the prin 2009 . some of the data presented herein were obtained at the w.m . keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and the national aeronautics and space administration . the observatory was made possible by the generous financial support of the w.m . keck foundation . the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of mauna kea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . herschel_-atlas is a project with _ herschel _ , which is an esa space observatory with science instruments provided by european - led principal investigator consortia and with important participation from nasa . the h - atlas website is http://www.h - atlas.org/. the us participants acknowledge support from the nasa _ herschel_science center / jpl . partly based on observations obtained with _ planck _ ( http://www.esa.int/planck ) , an esa science mission with instruments and contributions directly funded by esa member states , nasa , and canada . support for carma construction was derived from the gordon and betty moore foundation , the kenneth t. and eileen l. norris foundation , the james s. mcdonnell foundation , the associates of the california institute of technology , the university of chicago , the states of california , illinois , and maryland , and the national science foundation . ongoing carma development and operations are supported by the national science foundation under a cooperative agreement , and by the carma partner universities .	we present high - resolution maps of stars , dust , and molecular gas in a strongly lensed submillimeter galaxy ( smg ) at @xmath0 . hatlas j114637.9@xmath1001132 is selected from the _ herschel_-astrophysical terahertz large area survey ( h - atlas ) as a strong lens candidate mainly based on its unusually high 500 @xmath2 m flux density ( @xmath3300 mjy ) . it is the only high - redshift _ planck _ detection in the 130 deg@xmath4 h - atlas phase - i area . keck adaptive optics images reveal a quadruply imaged galaxy in the @xmath5-band while the submillimeter array and the jansky very large array show doubly imaged 880 @xmath2 m and co(1@xmath60 ) sources , indicating differentiated distributions of the various components in the galaxy . in the source plane , the stars reside in three major kpc - scale clumps extended over @xmath31.6 kpc , the dust in a compact ( @xmath31 kpc ) region @xmath33 kpc north of the stars , and the cold molecular gas in an extended ( @xmath37 kpc ) disk @xmath35 kpc northeast of the stars . the emission from the stars , dust , and gas are magnified by @xmath317 , @xmath38 , and @xmath37 times , respectively , by four lensing galaxies at @xmath7 . intrinsically , the lensed galaxy is a warm ( @xmath8 k ) , hyper - luminous ( @xmath9 @xmath10 ; sfr @xmath11 @xmath12 yr@xmath13 ) , gas - rich ( @xmath14 ) , young ( @xmath15 myr ) , and short - lived ( @xmath16 myr ) starburst . with physical properties similar to unlensed @xmath17 smgs , hatlas j114637.9@xmath1001132 offers a detailed view of a typical smg through a powerful cosmic microscope .
* sp - stm on the surface of fe@xmath0te bulk and thin films . * in spin - polarized scanning tunneling microscopy the tunneling current depends on the relative orientation of the tip magnetization and the local spin direction of the sample . the total tunneling current @xmath15 can be described by @xmath16 , where @xmath17 is the spin - averaged tunneling current and @xmath18 is the spin - polarized tunneling current which is proportional to the product of the spin - polarization @xmath19 of tip and @xmath20 of the sample , @xmath21 . here , @xmath22 is the angle between the spin direction of the tip and the sample . in general , fe - coated w - tips have a magnetization direction perpendicular to the tip axis and thus parallel to the surface of the sample . therefore , fe - coated w - tips exhibit sensitivity to the in - plane component of the sample magnetization . by applying an external magnetic field of about 1 t , the magnetization direction of the tip will be reoriented to the applied field direction @xcite . magnetic fields on the order of a tesla are orders of magnitude too weak to break the exchange interactions between the fe atoms in the bicollinear antiferromagnetic structure of fe@xmath0te . therefore , the application of external magnetic fields in different directions enables us to image the different spin components of the bicollinear antiferromagnetic structure by recording constant - current sp - stm images . in fig . [ fig : fete_simple_topo ] we give an introduction to both fe@xmath0te samples used in this work . the first investigated sample shown in figs . [ fig : fete_simple_topo](c , d ) is the surface of cleaved bulk fe@xmath0te ( @xmath23 ) . [ fig : fete_simple_topo](c ) displays a typical spin - resolved constant - current image of the bulk fe@xmath0te surface which shows clear atomic resolution of the te terminated surface ( a detailed analysis of the investigated spin contrast is discussed later on ) . compared to previously reported results @xcite a large area of the fe@xmath0te surface is free of excess fe atoms which is due to the annealing procedure as described in the methods section . this annealing procedure leads to the formation of fe clusters containing all the excess fe ( outside of the field of view of fig . [ fig : fete_simple_topo](c ) ) and large areas with no surface excess fe atoms in between these clusters . the surface is atomically flat but shows a small variation in the apparent height which is probably caused by excess fe atoms between the sub - surface layers . by calculating the fourier transform ( ft ) of fig . [ fig : fete_simple_topo](c ) , the lattice periodicity is displayed as bright bragg spots labeled as @xmath24 and @xmath25 in fig . [ fig : fete_simple_topo](d ) , were @xmath24 has a higher intensity than @xmath25 ( c.f . refs.@xcite ) . in addition to the atomic periodicity the constant - current image in fig . [ fig : fete_simple_topo](c ) shows the typical superstructure which has a periodicity of @xmath12 along the @xmath9-direction . the presence of this superstructure is visible as additional spots in the ft ( labeled with @xmath26 ) with a wave vector of @xmath26 = @xmath27 . overall the cleaved surface shows all the characteristics previously reported for fe@xmath0te ( cf . refs . @xcite ) . the second investigated sample is a thin fe@xmath0te film grown on bi@xmath3te@xmath4 ( figs . [ fig : fete_simple_topo](e - g ) ) . a large scale stm topography of the sample area is displayed in fig . [ fig : fete_simple_topo](e ) together with the height profile along a line shown in fig . [ fig : fete_simple_topo](f ) . from atomically resolved images taken on the different visible layers and their apparent heights , we deduce the structure of the layers as sketched in fig . [ fig : fete_simple_topo](g ) . the first uc thin layer of fete which has an apparent hight of @xmath283.5 ( fig . [ fig : fete_simple_topo](e ) ) is embedded into an incomplete bi@xmath3te@xmath4 quintuple layer . due to the interaction with the underlying substrate it exhibits a network of stripe - like dislocations leading to a rough surface ( not resolved in the large scale image of fig . [ fig : fete_simple_topo](e ) ) . on top of the embedded layer the growth of an additional fete layer has started and forms a second layer island with a height of 6.5 which is roughly equal to the @xmath29-axis lattice constant ( 6.26 ) @xcite ) of the bulk fe@xmath0te uc as shown in fig . [ fig : fete_simple_topo](a ) . this second layer has an atomically flat surface . in this work we focused on sp - stm measurements on top of the second layer islands indicated by the white arrow in fig . [ fig : fete_simple_topo](e ) . in fig . [ fig : fete_simple_topo](h ) an atomically resolved image of the surface on the second layer fe@xmath0te island is shown . we do not observe excess fe atoms on the surface , but we can not exclude that there is a small amount of excess fe atoms sitting in the van der waals gap in between the two fete layers . in addition to the atomic corrugation the characteristic @xmath13 superstructure is visible as indicated in the inset of fig . [ fig : fete_simple_topo](h ) , which is very similar to the one at the bulk fe@xmath0te surface shown in fig . [ fig : fete_simple_topo](c ) . this is also visible in the ft displayed in fig . [ fig : fete_simple_topo](i ) , which coincides with the ft pattern in fig . [ fig : fete_simple_topo](d ) ; the atomic lattice shows peaks of similar intensity ( labeled with @xmath24 and @xmath25 ) and also the @xmath13 superstructure peaks show similar intensities ( labeled with @xmath26 ) . therefore , the spin - resolved stm image of the second layer fe@xmath0te island exhibits a very similar spin structure as the surface of bulk fe@xmath0te . * magnetic field dependent sp - stm on the surface of bulk fe@xmath0te . * after introducing the two different sample systems used in this work we will now discuss on our results obtained with different orientations of the tip - magnetization starting with the surface of bulk fe@xmath0te and out - of plane tip sensitivity . figures [ fig : magnetic_contrast](a , b ) show constant - current maps of the same surface area obtained with an fe - coated tip . they were recorded in a magnetic field of 1 t with opposite out - of - plane field directions which forces the tip magnetization to point up and down . the two sp - stm images show the characteristic @xmath13 superstructure which is also visible by the @xmath26 peak displayed in the corresponding fts in fig . [ fig : magnetic_contrast](c , d ) . in contrast to different interpretations such as charge ordering phenomena @xcite , this additional superstructure has been attributed to a direct imaging of the bicollinear antiferromagentic order of fe@xmath0te by sp - stm @xcite . our experiments show that this is indeed the case and confirm that spin - polarized tunneling is the origin of the additional @xmath13 superstructure by applying external magnetic fields . by comparing the superstructure with the atomic lattice , a phase shift by one lattice unit is observed for opposite magnetic field directions ( cf . insets of figs . [ fig : magnetic_contrast](a , b ) ) . by subtracting the constant - current maps in fig . [ fig : magnetic_contrast](a ) from that in fig . [ fig : magnetic_contrast](b ) an image of the out - of plane components of the spin structure is obtained in fig . [ fig : magnetic_contrast](e ) which mainly shows a stripe pattern with a periodicity of @xmath13 along the @xmath9-direction . this is also reflected by the ft of the difference image in fig . [ fig : magnetic_contrast](f ) . here only the peak @xmath26 remains and the bragg peaks @xmath24 and @xmath25 related to the atomic lattice have vanished . this observation directly confirms the results from ref . @xcite and proves spin - polarized tunneling contrast due to the bicollinear antiferromagnetic spin structure of fe@xmath0te . the maximum spin contrast appears between every second fe lattice site located between two neighboring te sites . this can be explained by the fact that spin - polarized tunneling primarily results from the 3d states of fe being located below the top te layer . however , from the strong spin contrast we see using the out - of plane sensitive magnetic tip , we can additionally conclude , that the surface antiferromagnetic structure has a considerable out - of plane spin component . considering the magnetic structure of bulk fe@xmath0te as known from neutron diffraction@xcite depicted in fig . [ fig : fete_simple_topo](b ) this leads to the conclusion , that the surface spins are reoriented with respect to the corresponding bulk layers . in order to analyze the in - plane surface spin components of the bicollinear antiferromagnetic structure of fe@xmath0te , we performed experiments within a vector - magnet system which enables to rotate the tip magnetization direction within the film plane . for this purpose a magnetic field with a fixed amplitude was applied parallel to the surface and then stepwise rotated for each taken sp - stm image recorded . the results are shown in fig . [ fig : in - plane_neu ] for the magnetic field amplitude of 1 t. figure [ fig : in - plane_neu](a ) shows an overview of the atomically resolved fe@xmath0te surface with a magnetic field applied under an angle of @xmath30 relative to the given magnetic field coordinate system . on top of the atomic corrugation the strong @xmath13 superstructure is visible running continuously through the whole image from the bottom left to the top right . the blue square indicates the area where the investigations with rotated magnetic fields were performed and the red arrow indicates a defect used as a marker for the atomic - scale registry . upon rotating the magnetic field parallel to the surface the intensity of the @xmath13 superstructure was measured by taking spin - resolved constant - current maps for each field direction ( figs . [ fig : in - plane_neu](b - d ) , see full set of images in supplementary fig . the intensity of the superstructure was then extracted from the amplitude of the @xmath26-peak in the fts ( figs . [ fig : in - plane_neu](e - g ) ) . in addition to the @xmath26-peak the intensity of the bragg peaks @xmath24 @xmath25 were recorded as a reference in order to confirm , that the tip did not change during the full magnetic field sweep . these intensities are shown in fig . [ fig : in - plane_neu](i ) as a function of the magnetic field angle . in figs . [ fig : in - plane_neu](c , e ) , the tip magnetization direction is pointing in opposite directions within the surface plane . it is again apparent that the @xmath13 superstructure observed for both field directions exhibits a phase shift of one lattice unit . this is verified by calculating the difference of figs . [ fig : in - plane_neu](b , c ) shown in ( h ) , where only the magnetic signal of the @xmath13 superstructure remains . by comparing the intensities of the two bragg peaks @xmath24 and @xmath25 and the bicollinear @xmath26 peak extracted from the fts of these images shown in figs . [ fig : in - plane_neu](e , f ) we can conclude , that both sp - stm images exhibit the same amplitudes for the atomic corrugation and the @xmath13 superstructure , respectively as shown in the plot of fig . [ fig : in - plane_neu](i ) . in contrast , in the sp - stm image and its ft for a field direction of 84@xmath31 ( fig . [ fig : in - plane_neu](d , g ) the amplitude of the atomic corrugation remains at the same level but the amplitude of the @xmath13 superstructure is strongly reduced . overall the resulting angular dependence plotted in fig . [ fig : in - plane_neu](i ) reveals a periodic variation of the @xmath26 peak intensity with a periodicity of 180@xmath31 , while the intensities of the bragg peaks do not show significant changes . as shown by the fitted line , the @xmath26 intensity nicely follows a @xmath32-dependence which is expected for spin - polarized tunneling into an antiferromagnetic spin structure upon the rotation of the tip magnetization . the maximum of the intensity was found to be at -19@xmath31 . the corresponding in - plane spin direction of the bicollinear antiferromagnetic structure is indicated by the green arrow in fig . [ fig : in - plane_neu](a ) . we can thus conclude , that the spin direction of the surface layer bilcollinear antiferromagnetic structure deviates by 19@xmath31 from the @xmath10-direction , which is the spin direction of the bicollinear structure in bulk fe@xmath0te . * magnetic field dependent sp - stm on the surface of thin fe@xmath0te films on bi@xmath3te@xmath4 . * for comparison , we will now discuss the results obtained for a thin fe@xmath0te film grown on a bi@xmath3te@xmath4 substrate . all sp - stm data in fig . [ fig : mag_contrast_layer_neu ] were obtained in the same surface area on the top of a uc high fe@xmath0te island for magnetic fields of 2.5 t applied in opposite out - of - plane directions ( figs . [ fig : mag_contrast_layer_neu](a , b ) ) and for magnetic fields of 1.2 t applied in opposite in - plane directions ( figs . [ fig : mag_contrast_layer_neu](e , f ) ) . as for the measurements discussed above , the characteristic @xmath13 superstructure is observed in all four sp - stm data sets . the fts in the insets of figs . [ fig : mag_contrast_layer_neu](a , b ) and ( e , f ) also show the same @xmath26 pattern as for the bulk samples . by comparing the position of the maximum of the @xmath13 superstructure relative to the underlying te - lattice , e.g. at the position indicated by the red arrow , the fe@xmath0te thin film system also reveals a phase shift by one lattice constant upon inverting of the tip s magnetization direction in the out - of - plane as well as in the in - plane direction . this is also obvious from the difference images in figs . [ fig : mag_contrast_layer_neu](c , g ) reflecting the images of the out - of plane and in - plane components of the spin structure , respectively . therefore , we can conclude that the surface of the thin film fe@xmath0te on bi@xmath3te@xmath4 exhibits the same bicollinear antiferromagnetic spin structure as the bulk samples . additionally , in order to analyze the direction of the surface spins , the magnetic field dependence of the intensities of the @xmath26 peak and of the bragg peaks @xmath24 and @xmath25 were extracted for the out - of - plane and for the in - plane direction and are shown in figs . [ fig : mag_contrast_layer_neu](d , h ) , respectively ( full series of magnetic field dependent sp - stm images is given in the supplementary figs . s2 and s3 ) . upon increasing the magnetic field in the out - of - plane direction the intensity of @xmath26 starts changing since the magnetization direction of the tip is continuously rotated from the in - plane to the out - of - plane direction as shown in fig . [ fig : mag_contrast_layer_neu](d ) . for both the negative and the positive field direction the intensity of @xmath26 saturates when the tip magnetization is fully rotated into the out - of - plane direction . for the in - plane direction a similar behavior is observed as shown in fig . [ fig : mag_contrast_layer_neu](h ) . here , in zero magnetic field the initial tip magnetization has an unknown orientation with respect to the spin direction of the fe atoms in the surface fe@xmath0te layer . upon increasing the magnetic field , the tip continuously rotates into the direction of the magnetic field , where a strong magnetic contrast is deduced by the increase in the @xmath26 intensity . from the similar saturation values of the @xmath26 intensities in figs . [ fig : mag_contrast_layer_neu](d , h ) we can conclude , that the out - of plane component of the bicollinear antiferromagnetic spin structure in the surface layer of the fe@xmath0te thin film has a similar strength as the in - plane component . in summary , we provide a direct proof that spin - polarized tunneling is responsible for the observation of the @xmath13 superstructure in scanning tunneling microscopy on fe@xmath0te by using well defined spin - sensitive fe - coated w - tips . measurements under applied magnetic fields reveal that the @xmath13 superstructure can only be interpreted in terms of direct sp - stm imaging of the bicollinear afm order of the fe@xmath0te surface . this confirms previous sp - stm results with magnetically sensitive tips on fe@xmath0te @xcite and does not support the interpretation of a @xmath13 cdw - order discussed in ref . furthermore , we have found a strong contribution of the surface spin component in the out - of - plane direction along the c - axis and an in - plane component which deviates from the @xmath10-axis direction , for both the bulk fe@xmath0te samples and the thin fe@xmath0te films grown on bi@xmath3te@xmath4 . neutron scattering , which is sensitive to the bulk magnetization , revealed a dominant spin direction of the bicollinear antiferromagnetic structure along the @xmath33-axis direction @xcite . a tiny component of the magnetization along the @xmath34- and @xmath29-axes has been mainly attributed to local moments of excess iron atoms which are located in the van der waals gap between the fete layers @xcite . in contrast , our experiments indicate a strong component of the surface magnetization out of the @xmath33-axis direction , in favour of a reorientation of the ordered magnetic moments _ at the surface _ of fe@xmath0te . the central question is , thus , what drives this surface reorientation of the spin direction by keeping the overall bicollinear order . due to its layered crystal structure bulk fe@xmath0te has a quasi - two - dimensional electronic structure and the _ relative _ orientation of the spins within the bicollinear structure is mostly determined by the exchange interaction of the fe atoms within the @xmath34-@xmath35-plane of a given fe@xmath0te layer . this exchange interaction is relatively strong and largely unaffected by additional effects occurring at the surface such as lattice relaxations or charge redistribution . therefore , the bicollinear antiferromagnetic structure is preserved at the surface of fe@xmath0te . on the other hand , the _ absolute _ orientation of the magnetic moments is influenced by the magnetic anisotropy energy which has a much smaller energy scale . typically , the magnetic anisotropy is strongly influenced by electronic effects or lattice relaxations . based on these considerations , we propose that the observed reorientation of the spin structure at the surface of fe@xmath0te is induced by such surface effects . our findings might have implications for related sample systems . e.g. comparable spin - polarized stm studies could offer insight into the origin of the strong increased @xmath5 of the monolayer fese grown on srtio@xmath4 compared to bulk fese @xcite . although bulk fese does not show static magnetic order , spin fluctuations are known to exist and being coupled to superconducting pairing @xcite , which likewise might be strongly modified on the surface . in conclusion , we have shown that sp - stm experiments with fe - coated w - tips are well suited to investigate correlated electron systems such as fe@xmath0te with a complex electronic and magnetic structure . however , a complete characterization requires sp - stm experiments performed in 3d - vector - field systems offering field - dependent studies with arbitrary field orientation . * tip and sample preparation , experimental techniques . * high quality fe@xmath0te single crystals were synthesized using the flux method @xcite where the excess fe ratio @xmath7 was kept as low as possible . the measured composition of the crystals using single crystal x - ray diffraction resulted in @xmath36 . the samples were cleaved _ in situ _ at room temperature ( rt ) , and measured in ultra - high vacuum ( uhv ) with a background pressure better than @xmath37 mbar . for all sp - stm measurements the bulk samples were moderately annealed at 100 @xmath38c after cleaving which removes the surface excess iron and leads to an atomically flat surface . additionally , ultra - thin fe@xmath0te films were grown _ in situ _ on bi@xmath3te@xmath4 substrates . single crystals of bi@xmath3te@xmath4 were synthesized using a stockbarger method and were well characterized using arpes @xcite . fe - chalcogenide thin film preparation was carried out in a uhv system with a base pressure better than @xmath37 mbar . the bi@xmath3te@xmath4 crystals were cleaved _ in - situ _ under uhv conditions and fe@xmath0te thin films were prepared by depositing 0.5 - 1 monolayer ( ml ) fe onto a clean bi@xmath3te@xmath4 surface at rt followed by a 15 min annealing cycle at @xmath28300 @xmath38c . fe deposited on bi@xmath3te@xmath4 reacts with the substrate upon annealing , most likely via a substitutional process replacing bi by fe . this preparation was performed similar to the method described in ref . @xcite . here , the main difference in the growth is that the fe@xmath0te islands do not show a moir - pattern but a smooth atomically flat surface . the scanning tunneling microscopy experiments were performed in two home - built low temperature uhv - stm systems with out - of - plane magnetic fields up to 5 t and in - plane magnetic fields up to 2 t @xcite at the base temperature of 6.5 k. to prepare spin - sensitive tips , electrochemically etched w - tips were shortly heated to @xmath282000 @xmath38c ( flash ) and afterwards a thin fe film of @xmath2810 nm thickness was deposited onto the tip apex by e - beam deposition @xcite . all stm data were recorded in constant - current mode with a fixed bias voltage @xmath14 and a constant set - point for the tunneling current @xmath15 . the fourier transforms ( fts ) were calculated from the absolute value of the complex two - dimensional fast fourier transform which is proportional to the power spectral density . we are indebted to alexander balatsky , christopher triola and tim o. wehling for valuable discussions . t. h. acknowledges funding from project no . ha 6037/2 - 1 of the dfg . work at tulane is supported by the us doe under grant de - sc0014208 ( for material synthesis ) . u. r. s. , s. m. , a. k. and r. w. acknowledge funding via the erc advanced grand astonish ( no . l. c. , ph . h. and j. w. acknowledge funding through the dfg priority program spp1666 ( grant no . wi 3097/2 ) . we also thank the aarhus university research foundation for supporting the bi@xmath3te@xmath4 bulk crystal growth . m. b. , e. m. j. h. and b. b. i. acknowledges the financial support for center of materials crystallography ( cmc ) , funded by the danish national research foundation ( dnrf93 ) . t. h. acknowledges maciej bazarnik and lorenz schmidt for support in the stm labs . t. h. and j. w. designed the experiment . t. h. performed the sp - stm experiments on bulk fe@xmath0te . t. h. and u. r. s. have grown and characterized the thin film samples and performed the sp - stm experiments . t. h. , u. r. s. , s. m. , a. k. and l. c. analyzed the data . t. h. , j. w. and r. w. wrote the manuscript . j. h. and z. m. have grown the fe@xmath0te single crystals . e. m.j . h. , m. b. , b. b. i. and ph . h. have grown and characterized the bi@xmath3te@xmath4 samples and measured the composition of the fe@xmath0te single crystals . the authors declare no competing financial interests . 40ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1126/science.1243479 [ * * , ( ) ] link:\doibase 10.1126/science.1242996 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) `` , '' in @noop _ _ , ( , ) p. @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.88.094509 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) , @noop * * , ( ) , @noop * * , ( ) @noop * * , ( ) @noop * * , ( )	* establishing the relation between the ubiquitous antiferromagnetism in the non - superconducting parent compounds of unconventional superconductors and their superconducting phase is believed to be important for the understanding of the complex physics in these materials . going from the bulk systems to thin films strongly affects the phase diagram of unconventional superconductors . for fe@xmath0te , the parent compound of the fe@xmath0se@xmath1te@xmath2 superconductors , bulk sensitive neutron diffraction has revealed an in - plane oriented bicollinear antiferromagnetic structure . here , we show by spin - resolved scanning tunneling microscopy that on the surfaces of bulk fe@xmath0te , as well as on thin films grown on the topological insulator bi@xmath3te@xmath4 , the spin direction is canted both away from the surface plane and from the high - symmetry directions of the surface unit cell , while keeping the bicollinear magnetic structure . our results demonstrate that the magnetism at the fe - chalcogenide surface markedly deviates from a simple in - plane oriented bicollinear antiferromagnetic structure , which implies that the pairing at the surface of the related superconducting compounds might be different from that in the bulk . * the physics of many transition metal oxides ( tmo ) is dominated by strong electronic correlations which leads to exotic ground states and excitations with a dominant role of electronic charge and spin degrees of freedom . for example , for cuprate based high temperature superconductors ( htscs ) there have been growing evidence that charge and spin density wave - like states are inherent to these materials with a significant impact on the corresponding excitation spectrum @xcite . indeed , spin and charge ordering appear to be a key feature for understanding the physics of htscs @xcite . furthermore , among the correlated electron systems the recently discovered iron - based superconductors @xcite are of particular interest for the understanding of the interplay between superconductivity and magnetism . especially , the iron - chalcogenide system fe@xmath0se@xmath1te@xmath2 has gained high interest due to the surprisingly high @xmath5 superconductivity in single layer films of fese @xcite and its unique interplay between magnetism and superconductivity@xcite . for bulk single crystals fese exhibits a superconducting transition temperature @xmath5 of no higher than 10 k @xcite but in a single unit cell ( uc ) thick film grown on srtio@xmath4 @xmath5 can be increased above 100 k @xcite . this finding has spurred numerous investigations aiming at the understanding of how superconductivity evolves in transition metal chalcogenides from bulk to ultra - thin films . in particular , scanning tunneling microscopy ( stm ) with atomic - scale resolution has proven to be an indispensable tool for revealing the real - space electronic structure @xcite . however , most of the recent stm studies on tmo and iron - based superconductors have been focusing only on the charge degrees of freedom . with the recent developments in spin - polarized stm ( sp - stm ) @xcite , which accesses both charge and spin degrees of freedom on the atomic length scale , detailed investigations of tmo compounds and iron - based superconductors have now become possible . since fe@xmath0te exhibits double - stripe ( bicollinear ) antiferromagnetic ( afm ) order , contrasted with single - stripe afm order in iron pnictide superconductor parent compounds , the study on its mechanism of magnetic order has recently attracted a lot of attention . it provides a nonpolar charge - neutral te terminated surface upon cleaving @xcite . furthermore fe@xmath0te can be grown _ in situ _ by molecular beam epitaxy ( mbe ) with high quality @xcite . in bulk , fe@xmath0te is the nonsuperconducting parent compound of the transition metal chalcogenide system fe@xmath0se@xmath1te@xmath2@xcite and exhibits an afm bicollinear magnetic structure below its nel - temperature @xmath6 @xcite . depending on its excess iron concentration @xmath7 the nel - temperature varies from @xmath8 k @xcite . using neutron diffraction , it has been shown that the bulk fe spins are pointing along the diagonal of the fe - fe square network @xcite . the crystal structure and corresponding bicollinear afm order are schematically shown in figs . [ fig : fete_simple_topo](a , b ) . the magnetic phase transition of bulk fe@xmath0te is accompanied by a structural phase transition , with the structure changing from a tetragonal to a monoclinic phase for which the lattice constant @xmath9 is slightly larger than the lattice constant @xmath10 @xcite . the bicollinear structure itself shows a commensurate afm modulation along the @xmath9-direction with a wave length of @xmath11 and has a ferromagnetic ( fm ) coupling along the @xmath10-direction @xcite . stm data obtained for fe@xmath0te bulk and thin film samples have revealed atomic resolution of the te - terminated surface and a superstructure on top of the atomic corrugation having a periodicity of @xmath12 @xcite . the observation of this additional @xmath12 periodicity is mainly discussed in two contradictory models . on one hand , the interpretation for the @xmath13 superstructure is given in terms of a charge density wave ( cdw ) , where the charge density of the top te layer is modified by the spin density wave ( sdw ) of the underlying afm order of the fe - layer . the strong dependence on the bias voltage @xmath14 suggests a complex interplay between the charge and the magnetic order @xcite . in contrast to a common representation of spin and charge modulations in correlated electron systems @xcite , in this case the cdw would have the same periodicity as the sdw . opposed to this interpretation , experiments performed with magnetically sensitive tips , which were prepared by attaching excess fe atoms to the tip apex , give additional insight based on spin - polarized tunneling @xcite . the @xmath13 superstructure can unambiguously be assigned to a direct imaging of the bicollinear antiferromagnetic order of the underlying fe lattice . however , in previous studies the absolute orientation of the spins within the antiferromagnetic structure could not be revealed . in this work we investigated the fe@xmath0te surface of bulk samples and of thin fe@xmath0te films grown on bi@xmath3te@xmath4 by spin - polarized stm , revealing the bicollinear antiferromagnetic structure in both sample systems . moreover , by using fe coated w - tips in a vector - magnet system , we were able to rotate the tip magnetization direction both within the surface plane as well as perpendicular to the surface . since we thereby have access to the different components of the spin direction of the sample , we could unambiguously prove that the spin direction within the surface bicollinear antiferromagnetic structure is canted with respect to the crystallographic @xmath10-direction .
the subject of this paper is a self - consistent theory of specific soliton states that were originally predicted in ref . @xcite and reportedly observed experimentally.@xcite without any doubt , these states can be regarded as a hallmark of two - band superconductivity in mesoscopic doubly - connected samples . indeed , owing to the emergence of additional degrees of freedom of the order parameter , the nomenclature of topological objects in multiband superconductors is much richer than that in conventional single - band superconductors . in particular , ginzburg - landau equations describing two - band superconductivity in bulk samples admit topologically stable solutions ( with one - dimensional singularities of the order parameter ) that can be interpreted as vortices carrying fractional magnetic flux.@xcite in the absence of any interband coupling , these vortices are accompanied by a circulating neutral superfow associated with gradients of the interband phase difference . in the presence of `` josephson - type '' interband coupling , the neutral superflow generates static solitons of the sine - gordon type . in contrast to traditional abrikosov vortices in type - ii superconductors , the energy per unit length of these composite topological defects diverges at spatial infinity : hence they are thermodynamically metastable and difficult to create in bulk samples . however , solitons of the interband phase difference can exist by themselves in doubly - connected mesoscopic samples , when the formation of any magnetic vortices in the volume of the superconductor is prohibited energetically.@xcite moreover , soliton states in this case can be induced by an externally applied magnetic field , which makes them a convenient object of investigation . thus , experimental studies@xcite of the magnetic response of mesoscopic two - band superconducting rings reveal certain nontrivial features that , according to the authors of ref . @xcite , can be attributed to the creation of metastable soliton states . our research is largely motivated by the absence in current literature of any quantitative theoretical analysis of this pronounced feature of two - band superconductivity . ( unfortunately , the arguments of ref . @xcite and of the recent publications@xcite are mostly heuristic by nature . ) mathematically , the approach of this paper is based on a ginzburg - landau - type theory , which is a commonplace in theoretical studies of topological defects in two - band superconductors : see the next section . this means , of course , that we are restricted to the temperature range@xmath3 where @xmath4 is the critical temperature of the superconducting transition . , @xmath5 and @xmath6 obey conditions ( [ 1.3])-([1.6]).,scaledwidth=50.0% ] as to the physical object , we consider a two - band superconductor in the form of a straight , circular thin - walled cylinder , whose symmetry axis is the @xmath7 axis of cylindrical coordinates @xmath8 ( see fig . [ fig:1 ] ) . the constant external magnetic field @xmath9 is applied along the symmetry axis : @xmath10 . the length of the generatrix of the wall of the cylinder satisfies the condition@xmath11 which allows us to neglect end effects . the wall thickness @xmath12 and the average radius of the cylinder @xmath5 satisfy the following conditions:@xmath13@xmath14@xmath15@xmath16 where @xmath17 and @xmath18 are the `` coherence lengths '' in bands 1 and 2 , respectively , and @xmath19 is the weak - field penetration depth . exact definitions of @xmath17 , @xmath18 , and @xmath19 will be given in the next section ; however , the role of conditions ( [ 1.3])- ( [ 1.6 ] ) should be explained right now . thus , condition ( [ 1.3 ] ) precludes the formation of any magnetic vortices in the wall of the cylinder . condition ( [ 1.4 ] ) mainly simplifies mathematical consideration . in contrast , a combination of conditions ( [ 1.5 ] ) and ( [ 1.6 ] ) is of primary importance : taken together , these two conditions guarantee that self - induced magnetic fields are small and can be treated perturbatively . ( this fact justifies the definition `` mesoscopic cylinders '' in the title of the paper . ) moreover , as will be shown , the dimensionless quantity @xmath20 serves as a natural expansion parameter of the gibbs free energy . in order to carry out a rigorous analysis of thermodynamic stability of soliton states , we will have to evaluate the gibbs free energy _ exactly _ up to small terms of order @xmath21 , which implies the necessity of self - consistent evaluation of the vector potential up to first - order terms in @xmath20 . we conclude the formulation of the problem by specifying boundary conditions for soliton states . consider a two - component superconducting order parameter @xmath22 , where @xmath23 and @xmath24 . the double - connectedness of the cylinder is accounted for by the condition of single - valuedness of @xmath25.@xcite in particular , this condition requires that@xmath26 as to the phases @xmath27 and @xmath28 , the requirement is as follows:@xmath29 where @xmath30 is an arbitrary closed continuous contour that lies inside the wall of the cylinder and encircles the opening . it should be emphasized that there are _ no _ a priori reasons for setting @xmath31.@xcite as in the case of fractional magnetic vortices in bulk two - band superconductors,@xcite nontrivial topological states arise when @xmath32 . in the presence of interband coupling , they are of the soliton type . in section ii , we introduce the gibbs free - energy functional of the system and analyze its basic properties . in section iii , we derive a self - consistent expression for the electromagnetic gibbs free energy . soliton solutions are derived and thoroughly discussed in section iv . finally , in section v , we summarize the obtained results and make several concluding remarks . appendices a and b contain details of some mathematical calculations skipped over in the main text . in appendix c , we present several particular examples of soliton solutions . we begin by defining the gibbs free - energy functional of the system . in view of complete homogeneity along the @xmath7 axis and with the normal - state gibbs free energy being subtracted , it takes the following form:@xmath33 = l% % tcimacro{\dint \nolimits_{\sigma_{s}}}% % beginexpansion { \displaystyle\int\nolimits_{\sigma_{s } } } % endexpansion d^{2}\mathbf{r}\left [ \alpha_{1}\left\vert \psi_{1}\right\vert ^{2}% + \alpha_{2}\left\vert \psi_{2}\right\vert ^{2}+\frac{\beta_{1}}{2}\left\vert \psi_{1}\right\vert ^{4}+\frac{\beta_{2}}{2}\left\vert \psi_{2}\right\vert ^{4}\right . \nonumber\\ \left . + \frac{1}{2m_{1}}\left\vert \left ( -i\hbar\nabla-\frac{2e}% { c}\mathbf{a}\right ) \psi_{1}\right\vert ^{2}+\frac{1}{2m_{2}}\left\vert \left ( -i\hbar\nabla-\frac{2e}{c}\mathbf{a}\right ) \psi_{2}\right\vert ^{2}-\gamma\left ( \psi_{1}\psi_{2}^{\ast}+\psi_{1}^{\ast}\psi_{2}\right ) \right ] \nonumber\\ + \frac{l}{8\pi}% % tcimacro{\dint \nolimits_{\sigma_{s}+\sigma_{o}}}% % beginexpansion { \displaystyle\int\nolimits_{\sigma_{s}+\sigma_{o } } } % endexpansion d^{2}\mathbf{r}\left ( \mathbf{h}-\mathbf{h}\right ) ^{2}. \label{2.1}%\end{gathered}\ ] ] here , the coefficients @xmath34 and @xmath35 are positive constants , whereas @xmath36 and @xmath37 are temperature - dependent:@xmath38 moreover , the latter coefficients enter the definitions of the coherence lengths @xmath17 and @xmath18:@xmath39 the electron charge in ( [ 2.1 ] ) is @xmath40 ; the total vector potential @xmath41 defines the local magnetic field @xmath42:@xmath43 the parameter of interband coupling , @xmath44 , may have either sign . two - dimensional integration in the plane @xmath45 is carried out over the cross - section of the superconductor ( @xmath46 ) in the square - bracketed terms , and over the cross - sections of the superconductor and of the opening ( @xmath47 ) in the last ( magnetic ) term . a microscopic derivation of free - energy functionals of the type ( [ 2.1 ] ) was given in ref . @xcite for the case of clean two - band superconductors in the limit of small interband coupling . free - energy functionals of this type are employed in theoretical studies of different aspects of two - band superconductivity , such as , e.g. , topological defects,@xcite current - carrying states,@xcite the little - parks effect,@xcite surface energy,@xcite etc . it should be additionally noted that , for our specific geometry and @xmath48 , the functional ( [ 2.1 ] ) also applies to the description of a composite system consisting of two thin coaxial cylindrical films of single - band superconductors , josephson coupled via a parallel insulating layer,@xcite which is exactly the experimental set - up of ref . @xcite . to obtain the actual ( observable ) gibbs free energy , one has to minimize ( [ 2.1 ] ) with respect to @xmath25 , @xmath49 and @xmath41 under appropriate boundary conditions ; however , substantial simplifications can be made already at this stage . first , we notice that , by the symmetry of the problem , the amplitudes @xmath50 and @xmath51 do not depend on @xmath52 . moreover , they can not depend on @xmath53 , either . indeed , by virtue of condition ( [ 1.3 ] ) , any radial variations of @xmath50 and @xmath54 would give rise to free - energy terms that are by the factors @xmath55 and @xmath56 larger than the first and the second terms in ( [ 2.1 ] ) , respectively , which is energetically unfavorable.@xcite as a result , in equilibrium , the magnitudes @xmath50 and @xmath54 are functions of @xmath57 and @xmath58 only . consider now the kinetic - energy terms [ the first two terms in the second line of ( [ 2.1 ] ) ] . the ratio of these terms to the first and the second terms in ( [ 2.1 ] ) , respectively , is at most of order @xmath59 and @xmath60 , where @xmath61 is the external flux , and @xmath62 is the flux quantum . owing to condition ( [ 1.4 ] ) , for sufficiently weak external fields , @xmath63 . ( compare with the consideration of flux quantization in singe - band - superconducting cylinders.@xcite ) in this field range , we can set @xmath64 and @xmath65 , where @xmath66 and @xmath67 satisfy the equilibrium conditions for an unperturbed two - band superconductor : @xmath68 one can readily obtain a good approximate solution to ( [ 2.6]).@xcite however , it is of no interest in the context of the soliton problem . we only note that the critical temperature , derived from ( [ 2.6 ] ) , is @xmath69 $ ] . in light of these simplifications , it is reasonable to consider the weak - field penetration depth@xcite@xmath70 and to define intraband `` penetration depths''@xcite@xmath71 ( for the above - mentioned composite , josephson - coupled system , the quantities @xmath72 and @xmath73 have direct physical meaning . ) we also introduce new , functionally independent phase variables @xmath74 and @xmath75:@xcite@xmath76 using definitions ( [ 2.5 ] ) and ( [ 2.8])-([2.11 ] ) , we obtain the reduced gibbs free - energy functional in the following form:@xmath77 = f_{s0}+g_{em}\left [ \chi\,,\mathbf{a};\mathbf{h}\right ] + f_{sol}\left [ \phi\right ] . \label{2.12}%\ ] ] here , the first term is the free energy of the unperturbed superconducting cylinder:@xcite@xmath78 the second term is the electromagnetic gibbs free - energy functional:@xmath79 = \frac{\phi_{0}^{2}}% { 32\pi^{3}}\frac{l}{\lambda^{2}}% % tcimacro{\dint \nolimits_{\sigma_{s}}}% % beginexpansion { \displaystyle\int\nolimits_{\sigma_{s } } } % endexpansion d^{2}\mathbf{r}\left ( \nabla\chi-\frac{2e}{\hbar c}\mathbf{a}\right ) ^{2}+\frac{l}{8\pi}% % tcimacro{\dint \nolimits_{\sigma_{s}+\sigma_{o}}}% % beginexpansion { \displaystyle\int\nolimits_{\sigma_{s}+\sigma_{o } } } % endexpansion d^{2}\mathbf{r}\left ( \mathbf{h}-\mathbf{h}\right ) ^{2 } , \label{2.14}%\ ] ] with the first term on the right - hand side of ( [ 2.14 ] ) being the kinetic - energy functional of the supercurrent . finally , the last term in ( [ 2.12 ] ) is@xmath80 = \frac{\phi_{0}^{2}}{32\pi^{3}}\frac{l}% { \lambda^{2}}c_{1}c_{2}% % tcimacro{\dint \nolimits_{\sigma_{s}}}% % beginexpansion { \displaystyle\int\nolimits_{\sigma_{s } } } % endexpansion d^{2}\mathbf{r}\left [ \left ( \nabla\phi\right ) ^{2}+\frac{2}{l^{2}}\left ( 1-\text{sgn~}\gamma\cos\phi\right ) \right ] ; \label{2.15}\\ l^{2}\equiv\frac{\phi_{0}^{2}}{32\pi^{3}}\frac{1}{\lambda^{2}}\frac{c_{1}% c_{2}}{\left\vert \gamma\right\vert \left\vert \psi_{1}\right\vert _ { 0}\left\vert \psi_{2}\right\vert _ { 0}},\nonumber\end{gathered}\ ] ] where sgn @xmath81 is the sign function . the term ( [ 2.15 ] ) should be interpreted as the soliton self - energy functional . indeed , when @xmath31 in ( [ 1.8 ] ) , we have@xcite either @xmath82 mod @xmath83 ( for @xmath48 ) or @xmath84 mod @xmath83 ( for @xmath85 ) , and this term vanishes identically . our task now is to minimize ( [ 2.12 ] ) with respect to @xmath86 , @xmath75 and @xmath41 . as the phase variable @xmath86 is not coupled to the vector potential @xmath41 , this procedure can be performed in two separate steps . the minimization of the electromagnetic functional ( [ 2.14 ] ) reduces to evaluation of the stationarity condition @xmath87 , or@xmath88 indeed , in view of quadratic nature of ( [ 2.14 ] ) , solutions to ( [ 2.17 ] ) are automatically minimizers of this functional ( i.e. , @xmath89 at these solutions ) . variation with respect to @xmath41 yields ampre s law@xmath90 with@xmath91 being the supercurrent density [ @xmath92 by symmetry ] , and the boundary condition@xmath93 ( this boundary condition should , of course , be complemented by the conditions of continuity of @xmath41 and @xmath42 at @xmath94 and the condition of regularity of @xmath41 at the origin . ) variation with respect to @xmath95 , under the condition of single - valuedness of variations @xmath96 , just yields the current - conservation law@xmath97 and the single - valuedness condition@xmath98 [ this boundary condition should be complemented by a condition on @xmath75 resulting from ( [ 1.8 ] ) . ] the problem of finding @xmath41 and @xmath75 is still sub - definite , because we have not so far fixed the gauge . as the @xmath7 component of the vector potential drops out of the definition of @xmath42 [ see ( [ 2.4 ] ) ] , it is equal to an arbitrary constant , and we set @xmath99 . the @xmath53 component of the vector potential can be eliminated by the gauge transformation@xmath100 in this particular gauge,@xmath101 and @xmath75 does not depend on @xmath53 ( @xmath102 ) . using ( [ 1.8 ] ) , ( [ 2.11 ] ) , ( [ 2.22 ] ) and ( [ 2.23 ] ) , we arrive at a well - posed boundary - value problem,@xmath103 whose solution is@xmath104 with @xmath105 being an arbitrary constant . the boundary - value problem for the vector potential now takes the form@xmath106 = 0,\quad r\in\left ( 0,r_{1}\right ) ; \nonumber\\ \frac{d}{dr}\left [ \frac{1}{r}\frac{d}{dr}\left ( ra\right ) \right ] = \frac{1}{\lambda^{2}}\left [ a+\frac{\phi_{0}}{2\pi}q\left ( n_{1}% , n_{2}\right ) \right ] , \quad r\in\left ( r_{1},r_{2}\right ) ; \nonumber\\ q\left ( n_{1},n_{2}\right ) \equiv n_{1}c_{1}+n_{2}c_{2};\label{2.28}\\ \left . \left\vert a\right\vert \right\vert _ { r=0}<\infty,\,\left . { r = r_{1}-0}=\left . a\right\vert _ { r = r_{1}+0},\,\left . \frac { 1}{r}\frac{d}{dr}\left ( ra\right ) \right\vert _ { r = r_{1}-0}=\left . \frac { 1}{r}\frac{d}{dr}\left ( ra\right ) \right\vert _ { r = r_{1}+0},\nonumber\\ \left . \frac{1}{r}\frac{d}{dr}\left ( ra\right ) \right\vert _ { r = r_{2}% } = h.\nonumber\end{gathered}\ ] ] this boundary - value problem admits an exact solution : it is presented in appendix a. however , to obtain a second - order expansion of the electromagnetic gibbs free energy in terms of the small parameter @xmath20 ( see introduction ) , we need only first - order expansions of @xmath107 and @xmath108 . they are as follows:@xmath109 \varepsilon,\quad r\in\left [ 0,r_{1}\right ] ; \nonumber\\ & = \frac{r}{2}h-\frac{r_{1}}{2}\left [ \frac{\phi_{0}}{\pi r_{1}^{2}}q\left ( n_{1},n_{2}\right ) + h\right ] \varepsilon,\quad r\in\left ( r_{1}% , r_{2}\right ] ; \label{2.29}%\end{aligned}\]]@xmath110 \varepsilon,\quad r\in\left [ 0,r_{1}\right ] ; \nonumber\\ & = h-\frac{r_{2}-r}{r_{2}-r_{1}}\left [ \frac{\phi_{0}}{\pi r_{1}^{2}% } q\left ( n_{1},n_{2}\right ) + h\right ] \varepsilon,\quad r\in\left ( r_{1},r_{2}\right ] . \label{2.30}%\end{aligned}\ ] ] [ the fact that expressions ( [ 2.29 ] ) and ( [ 2.30 ] ) on the interval @xmath111 are _ not _ related to each other by equation ( [ 2.26 ] ) should not cause any confusion : to ensure the fulfillment of ( [ 2.26 ] ) , we would have to continue the expansion of ( [ a4 ] ) up to small terms of order @xmath112 and @xmath113 . ] bearing in mind that in integral physical quantities any difference between @xmath114 , @xmath115 and @xmath5 should be neglected [ see ( [ a8 ] ) ] , by use of equations ( [ 2.30 ] ) and ( [ 2 - 19 ] ) we immediately obtain expressions for the total flux @xmath116 and the total supercurrent @xmath117:@xmath118 \varepsilon ; \label{2.31}%\]]@xmath119 here , @xmath61 and @xmath120 are the external and self - induced flux , respectively ; @xmath1 is the magnetic inductance of the cylinder ( or self - inductance).@xcite on substitution of relations ( [ 2.27 ] ) , ( [ 2.29 ] ) and ( [ 2.30 ] ) into ( [ 2.14 ] ) , we get the electromagnetic gibbs free energy of the state parameterized by topological numbers @xmath121 and @xmath122:@xmath123 ^{2}}{2\mathcal{l}_{m}}\varepsilon \left ( 1 - 2\varepsilon\right ) + \frac{\left [ \phi_{0}q\left ( n_{1}% , n_{2}\right ) + \phi_{h}\right ] ^{2}}{2\mathcal{l}_{m}}\varepsilon ^{2}\nonumber\\ = \frac{\phi_{0}^{2}\varepsilon}{2\mathcal{l}_{m}}\left [ q\left ( n_{1}% , n_{2}\right ) + f_{h}\right ] ^{2}\left ( 1-\varepsilon\right ) , \quad f_{h}\equiv\frac{\phi_{h}}{\phi_{0}}. \label{2.33}%\end{gathered}\ ] ] ( note that the first term in the first line of this equation is the kinetic energy of the supercurrent , whereas the second term in the same line is the magnetic gibbs free energy . ) to facilitate an analysis of soliton states , we should transform ( [ 2.33 ] ) to a more convenient form . first , instead of parameterization by @xmath121 and @xmath122 , we introduce parameterization by @xmath121 and the soliton number @xmath124 . from now on , we assume ( without any loss of generality ) that @xmath125 . relation ( [ 2.33 ] ) is rewritten as follows:@xmath126 from a thermodynamic point of view , of interest is the minimum of ( [ 2.34 ] ) for given @xmath127.therefore , relation ( [ 2.34 ] ) should be minimized with respect to @xmath121 and sng@xmath128 . with this in mind , we introduce two discontinuous functions , a step function @xmath129 and a periodic function @xmath130 , via the definitions@xmath131{c}% \left [ x\right ] , \quad0\leq\left\ { x\right\ } \leq\frac{1}{2};\\ \left [ x\right ] + 1,\quad\frac{1}{2}<\left\ { x\right\ } < 1 , \end{array } \right . \label{2.35}%\ ] ] and@xmath132{c}% \left\ { x\right\ } , \quad0\leq\left\ { x\right\ } \leq\frac{1}{2};\\ -1+\left\ { x\right\ } , \quad\frac{1}{2}<\left\ { x\right\ } < 1 , \end{array } \right . \label{2.36}%\ ] ] where @xmath133 $ ] and @xmath134 are the integer and fractional parts of @xmath81 , respectively . given that @xmath135 and @xmath136 can now be represented as @xmath137 and @xmath138sgn @xmath139 $ ] , respectively , the result of the minimization is@xmath140 ^{2}\left ( 1-\varepsilon \right ) . \label{2.37}%\end{gathered}\ ] ] the variation of ( [ 2.15 ] ) with respect to @xmath86 , under the condition of single - valuedness of variations @xmath141 , yields a static two - dimensional sine - gordon equation in polar coordinates,@xmath142 and the boundary conditions@xmath143 [ these boundary conditions should , of course , be complemented by a condition on @xmath86 resulting from ( [ 1.8 ] ) . ] however , equation ( [ 3.1 ] ) , in its exact form , by far exceeds the accuracy of our calculations in the previous section [ see expressions ( [ 2.29 ] ) , ( [ 2.30 ] ) and relations ( [ a8 ] ) ] . discarding in ( [ 3.1 ] ) terms of order @xmath112 and @xmath144 , we arrive at a two - dimensional sine - gordon equation in `` cartesian coordinates'':@xmath145 solutions to ( [ 3.3 ] ) , minimizing the functional ( [ 2.15 ] ) , should not depend on @xmath146 for symmetry reasons [ which , of course , is compatible with boundary conditions ( [ 3.2 ] ) ] . thus , the phase @xmath147 satisfies the following boundary - value problem:@xmath148 the solution of ( [ 3.4 ] ) is straightforward:@xcite@xmath149 where am @xmath150 is the elliptic amplitude,@xcite @xmath151 is the complete elliptic integral of the first kind,@xcite @xmath152 are arbitrary constants , and @xmath153 ( @xmath154 ) satisfy the equations@xmath155 particular examples of solutions ( [ 3.5 ] ) that possess asymptotics in terms of elementary functions are relegated to appendix c. nonetheless , the very special class of _ exact _ elementary solutions is worth being presented here : namely , the _ non_-soliton nontrivial topological solutions corresponding to the physical case of the absence of interband coupling ( @xmath156 ) . these solutions can be obtained from ( [ 3.5 ] ) by the limit procedure@xmath157 and they have the general form@xmath158 they are necessarily minimizers of ( [ 2.15 ] ) [ i.e. , at these solutions @xmath159 , because the functional ( [ 2.15 ] ) is quadratic in the case @xmath156 ] , and their self - energy is@xmath161 if @xmath162 , the functional ( [ 2.15 ] ) is non - quadratic , and we should analyze the second variation of ( [ 2.15 ] ) in more detail . to this end,@xcite we turn to the sturm - liouville problem@xmath163 where @xmath164 is a given solution from the set ( [ 3.5 ] ) . as shown in refs . @xcite,@xmath165 where @xmath166 is lowest eigenvalue of the problem ( [ 3.10 ] ) . in our case , both @xmath166 and the corresponding eigenfunction @xmath167 can be readily found:@xmath168 where dn@xmath169.@xcite this means that @xmath170 , and soliton states turn out to be _ indifferently _ stable states . indeed , the zero value of @xmath166 should be attributed to the existence of a zero - frequency `` rotational mode '' ( by analogy with the well - known@xcite translational mode in quantum field theories ) that restores rotational symmetry broken by the formation of solitons . to prove this , consider a small variation of @xmath171 induced by a small variation of the constant of integration @xmath152:@xmath172 from the above , we see that @xmath173 . now that local stability of soliton solutions is established , we proceed with a discussion of soliton self - energy . it is obtained by the substitution of solutions ( [ 3.5 ] ) into ( [ 2.15 ] ) and has the form@xmath174 , \label{3.12}%\ ] ] where @xmath175 is the complete elliptic integral of the second kind.@xcite first , we note that the constants of integration @xmath152 that figure in ( [ 3.5 ] ) drop out of the right - hand side of ( [ 3.12 ] ) , as they should . the self - energy does not depend on the sign of @xmath44 and of @xmath176 , either . by considering ( formally ) @xmath177 as a continuous variable , we get@xmath178 which means that @xmath179 increases monotonically with an increase in @xmath177 , as could be expected . however , in contrast to the case @xmath156 [ see ( [ 3.9 ] ) ] , the growth of @xmath179 is slower than @xmath180 , because@xmath181 = -\frac{\phi_{0}% ^{2}\varepsilon}{\mathcal{l}_{m}}\frac{4r}{\pi\left\vert n\right\vert ^{2}% l}c_{1}c_{2}\frac{e\left ( k_{n}\right ) -\left ( 1-k_{n}^{2}\right ) k\left ( k_{n}\right ) } { k_{n}}<0.\ ] ] given that@xmath182 } { k_{n}\left ( 1-k_{n}^{2}\right ) } > 0,\quad k_{n}\in\left ( 0,1\right ) , \ ] ] the self - energy increases monotonically with an increase in @xmath153 on the whole interval @xmath183 . the minimal value of ( [ 3.12 ] ) is achieved at @xmath184 and is given by ( [ 3.9 ] ) . in view of the relation@xmath185 the self - energy decreases monotonically with an increase in @xmath186 ( for a given @xmath177 ) . [ in other words , @xmath187 is an _ increasing _ function of the interband coupling parameter @xmath188 : see the definition of @xmath189 in ( [ 2.15 ] ) . ] according to ( [ 2.12 ] ) , ( [ 2.37 ] ) and ( [ 3.12 ] ) the minimal gibbs free energy of soliton states with a given @xmath177 in the field @xmath58 can be represented as follows:@xmath190 ^{2}\left ( 1-\varepsilon\right ) \right . \nonumber\\ \left . + \frac{4\left\vert n\right\vert ^{2}}{\pi^{2}}c_{1}c_{2}k\left ( k_{n}\right ) \left [ 2e\left ( k_{n}\right ) -\left ( 1-k_{n}^{2}\right ) k\left ( k_{n}\right ) \right ] \right ] . \label{3.14}%\end{gathered}\ ] ] to analyze thermodynamic stability of soliton solutions , we should compare expression ( [ 3.14 ] ) for @xmath191 with the gibbs free energy of the states with @xmath192:@xmath193 with this in mind , we first note that , for @xmath191 , the energy @xmath194 increases monotonically with an increase in @xmath177 : see appendix b for a proof . [ contrary to what may seem , this fact is by no means obvious , because the electromagnetic term in ( [ 3.14 ] ) may _ decrease _ with an increase in @xmath177 . ] furthermore , since expression ( [ 3.9 ] ) provides the greatest lower bound for soliton self - energies , we can restrict ourselves to the case @xmath195 and @xmath184 . bearing in mind that @xmath196 $ ] by assumption ( see the end of section iii ) , we arrive at the following important inequalities:@xmath197 > 0 . \label{3.16}%\end{gathered}\ ] ] the above inequalities clearly demonstrate thermodynamic metastability of soliton states and bring to light certain subtle physical points . in particular,@xmath198 in contrast,@xmath199 which shows that @xmath200 can be much smaller than the magnetic gibbs free energy of the zero - soliton states [ see ( [ 2.33 ] ) ] , provided that @xmath201 [ i.e. , when @xmath202 , see ( [ 2.11 ] ) ] and @xmath203 ( weak interband coupling ) . @xmath204{cc}% \includegraphics[width=0.5\textwidth]{fig_2_1.eps } & \includegraphics[width=0.5\textwidth]{fig_2_2.eps}\\ \includegraphics[width=0.5\textwidth]{fig_2_3.eps } & \includegraphics[width=0.5\textwidth]{fig_2_4.eps } \end{array } $ ] in fig . [ fig:2 ] , we plot the gibbs free energy of several different topological states @xmath205 ( @xmath206 ) . thermodynamically stable zero - soliton states are denoted by thick solid lines . minima of the gibbs free energy of soliton states represent the soliton self - energy and occur when the self - induced flux @xmath120 compensates for the external flux @xmath61 , i.e. , when@xmath207 [ see relations ( [ 2.31 ] ) , ( [ 2.34 ] ) and ( [ 2.37 ] ) ] . in the very special case , when @xmath208 and @xmath209 , no flux is induced ( @xmath210 ) , and minima of the soliton gibbs free energy occur at @xmath211 . summarizing , we have presented ( in the framework of the ginzburg - landau approach ) a self - consistent theory of specific soliton states that constitute a distinctive feature of two - band superconductivity in mesoscopic multiply - connected samples . although our mathematical consideration concerns the concrete geometry of fig . 1 , the final results can be expressed in terms of the magnetic and kinetic inductance ( see ref . @xcite ) and , therefore , should apply to a much wider class of structures . this allows us to make several generalizing remarks . as the predicted fractional magnetic vortices in bulk two - band superconductors,@xcite the soliton states considered here prove to be thermodynamically metastable . however , the minimal energy gap between the lowest - lying single - soliton states and thermodynamically stable zero - soliton states can be much smaller than the magnetic gibbs free energy of the latter states , provided that the intraband `` penetration depths '' ( [ 2.9 ] ) differ substantially and the interband coupling is weak . ( in order to establish this important physical fact , we had to evaluate self - consistantly the vector potential . the results of this evaluation may be of interest in themselves . ) our consideration encompasses in a natural way the case of superconducting josephson - coupled bilayer structures studied experimentally in ref . our conclusion that the self - energy of soliton states increases monotonically with an increase in the strength of interband coupling qualitatively agrees with the observations reported therein . furthermore , as a particular limit , our consideration contains the case of zero interband coupling . in view of the recently discussed possibility of independent superconductivity of electrons and protons in a liquid metallic state of hydrogen,@xcite some of our results may find application in this situation as well . finally , the exact soliton solutions derived in this paper should be compared with the exact soliton solutions representing equilibrium josephson vortices in a superconducting tunnel junction.@xcite in particular , josephson vortices are pinned in their equilibrium positions owing to interaction with the edges of the junction . in contrast , in rotationally symmetric doubly - connected two - band superconductors , soliton positions are not fixed , which gives rise to a specific zero - frequency rotational mode . however , any defects that break rotational symmetry must cause soliton pinning . the effect of this pinning requires a separate discussion . in conclusion , we hope that our paper will stimulate further experimental and theoretical studies of the intriguing phenomenon of soliton states in two - band superconductors . the exact solution to ( [ 2.28 ] ) has the following form:@xmath212 ; \nonumber\\ = -\frac{\phi_{0}}{2\pi r}q\left ( n_{1},n_{2}\right ) + c_{1}i_{1}\left ( \frac{r}{\lambda}\right ) + c_{2}k_{1}\left ( \frac{r}{\lambda}\right ) , \quad r\in\left ( r_{1},r_{2}\right ] ; \label{a1}%\end{gathered}\]]@xmath213 ; \nonumber\\ = \frac{c_{1}}{\lambda}i_{0}\left ( \frac{r}{\lambda}\right ) -\frac{c_{2}% } { \lambda}k_{0}\left ( \frac{r}{\lambda}\right ) , \quad r\in\left ( r_{1},r_{2}\right ] ; \label{a2}%\end{gathered}\]]@xmath214 } ; \label{a3}%\]]@xmath215 } { r_{1}f_{1}\left ( \frac{r_{2}}{\lambda},\frac{r_{1}}{\lambda}\right ) + 2\lambda f_{2}\left ( \frac{r_{2}}{\lambda},\frac{r_{1}}{\lambda}\right ) } + \frac{\lambda\phi_{0}q\left ( n_{1},n_{2}\right ) k_{0}\left ( \frac{r_{2}% } { \lambda}\right ) } { \pi r_{1}\left [ r_{1}f_{1}\left ( \frac{r_{2}}{\lambda } , \frac{r_{1}}{\lambda}\right ) + 2\lambda f_{2}\left ( \frac{r_{2}}{\lambda } , \frac{r_{1}}{\lambda}\right ) \right ] } , \]]@xmath216 } { r_{1}f_{1}\left ( \frac{r_{2}}{\lambda},\frac{r_{1}}{\lambda}\right ) + 2\lambda f_{2}\left ( \frac{r_{2}}{\lambda},\frac{r_{1}}{\lambda}\right ) } + \frac{\lambda\phi_{0}q\left ( n_{1},n_{2}\right ) i_{0}\left ( \frac{r_{2}% } { \lambda}\right ) } { \pi r_{1}\left [ r_{1}f_{1}\left ( \frac{r_{2}}{\lambda } , \frac{r_{1}}{\lambda}\right ) + 2\lambda f_{2}\left ( \frac{r_{2}}{\lambda } , \frac{r_{1}}{\lambda}\right ) \right ] } ; \]]@xmath217 here , @xmath218 and @xmath219 are modified bessel functions of order @xmath220;@xcite @xmath221 is the constant magnetic field in the opening . expressions ( [ a1])-([a3 ] ) are greatly simplified under condition ( [ 1.5]):@xmath212 ; \nonumber\\ = -\frac{\phi_{0}q\left ( n_{1},n_{2}\right ) } { 2\pi r}\frac{\sinh\frac { d}{\lambda}+\frac{2\lambda}{r_{1}}\left ( \cosh\frac{d}{\lambda}-\sqrt { \frac{r}{r_{1}}}\cosh\frac{r_{2}-r}{\lambda}\right ) } { \sinh\frac{d}{\lambda } + \frac{2\lambda}{r_{1}}\cosh\frac{d}{\lambda}}\nonumber\\ + \lambda h\sqrt{\frac{r_{2}}{r}}\frac{\cosh\frac{r - r_{1}}{\lambda}% + \frac{2\lambda}{r_{1}}\sinh\frac{r - r_{1}}{\lambda}}{\sinh\frac{d}{\lambda } + \frac{2\lambda}{r_{1}}\cosh\frac{d}{\lambda}},\quad r\in\left ( r_{1}% , r_{2}\right ] ; \label{a4}%\end{gathered}\]]@xmath213 ; \nonumber\\ = -\frac{\phi_{0}q\left ( n_{1},n_{2}\right ) } { \pi r_{1}\sqrt{rr_{1}}}% \frac{\sinh\frac{r_{2}-r}{\lambda}}{\sinh\frac{d}{\lambda}+\frac{2\lambda } { r_{1}}\cosh\frac{d}{\lambda}}\nonumber\\ + h\sqrt{\frac{r_{2}}{r}}\frac{\sinh\frac{r - r_{1}}{\lambda}+\frac{2\lambda } { r_{1}}\cosh\frac{r - r_{1}}{\lambda}}{\sinh\frac{d}{\lambda}+\frac{2\lambda } { r_{1}}\cosh\frac{d}{\lambda}},\quad r\in\left ( r_{1},r_{2}\right ] ; \label{a5}%\end{gathered}\]]@xmath222 expressions ( [ a4])-([a6 ] ) should be compared with analogous expressions for a single - band - superconducting cylinder.@xcite as can be easily seen , the cylinder exhibits a considerable meissner effect under the conditions@xmath223 in contrast , in the opposite case , when condition ( [ 1.6 ] ) is fulfilled , the meissner effect is small , and expressions ( [ a4])-([a6 ] ) can be readily expanded up to first - order terms in @xmath20 . taking into account a hierarchy of the small parameters of the problem,@xmath224 we arrive at the first - order expressions ( [ 2.29 ] ) and ( [ 2.30 ] ) . consider the expression@xmath226 \nonumber\\ = -\left\vert \theta\left ( \left\vert n\right\vert c_{2}\right ) \right\vert \left [ 2\left\vert \theta\left ( f_{h}\right ) \right\vert -\left\vert \theta\left ( \left\vert n\right\vert c_{2}\right ) \right\vert \right ] \left ( 1-\varepsilon\right ) + \frac{4\left\vert n\right\vert ^{2}}{\pi^{2}% } c_{1}c_{2}k\left ( k_{n}\right ) \left [ 2e\left ( k_{n}\right ) -\left ( 1-k_{n}^{2}\right ) k\left ( k_{n}\right ) \right ] \label{b1}%\end{gathered}\ ] ] that follows directly from ( [ 3.14 ] ) . our task is to prove that the right - hand side of ( [ b1 ] ) for @xmath227 is larger than for @xmath195 . given that @xmath228 $ ] by assumption ( see the end of section iii ) , it is sufficient to provide a proof for @xmath229 + 1 $ ] , where @xmath230 $ ] is the integer part of @xmath231 . indeed , the first term on the right - hand side of ( [ b1 ] ) satisfies the inequality@xmath232 \left ( 1-\varepsilon\right ) \right\vert \leq\frac{1}{4}\left ( 1-\varepsilon\right ) , \ ] ] whereas@xmath233 \geq1.\ ] ] therefore , for @xmath234 + 1 $ ] , when @xmath235 , any possible decrease in the first term on the right - hand side of ( [ b1 ] ) due to an increase in @xmath177 can not compensate for an incurred increase in the second term . for @xmath236 + 1 $ ] , there holds the relation @xmath237 , and ( [ b1 ] ) becomes@xmath238 + \left\vert n\right\vert ^{2}c_{1}c_{2}\left [ \frac{4k\left ( k_{n}\right ) } { \pi^{2}}\left [ 2e\left ( k_{n}\right ) -\left ( 1-k_{n}^{2}\right ) k\left ( k_{n}\right ) \right ] -1\right ] \\ = \left\vert n\right\vert \frac{2\mathcal{l}_{m}}{\phi_{0}^{2}\varepsilon } \delta g\left ( 1;f_{h}\right ) + c_{2}\left\vert n\right\vert \left ( \left\vert n\right\vert -1\right ) \left [ \left ( 1-c_{2}\varepsilon\right ) + \frac{4c_{1}k\left ( k_{n}\right ) } { \pi^{2}}\left [ 2e\left ( k_{n}\right ) -\left ( 1-k_{n}^{2}\right ) k\left ( k_{n}\right ) \right ] -c_{1}\right ] \\ > \frac{2\mathcal{l}_{m}}{\phi_{0}^{2}\varepsilon}\delta g\left ( 1;f_{h}\right ) , \end{gathered}\ ] ] which was to be proved . in fig . [ fig:3 ] , we present several different soliton solutions obtained numerically . ( for greater clarity , we plot the derivatives @xmath239 . ) however , in two limiting cases soliton solutions possess asymptotics in terms of elementary functions . thus , for @xmath240 , we have:@xcite@xmath241 . \label{c1}%\ ] ] the self - energy of soliton solutions ( [ c1 ] ) is@xmath242 [ notice that expression ( [ c2 ] ) clearly illustrates the general features of the self - energy of soliton solutions established in section iv . ] in the opposite limiting case , when @xmath243 , asymptotics can be derived only for the single - soliton solutions ( @xmath244 ) . fixing the constants of integration by the condition @xmath245 , we get:@xmath246 .\label{c3}%\ ] ] the self - energy of these solutions is@xmath247 .\label{c4}%\ ] ] @xmath204{cc}% \includegraphics[width=0.5\textwidth]{fig_3_1.eps } & \includegraphics[width=0.5\textwidth]{fig_3_2.eps}\\ \includegraphics[width=0.5\textwidth]{fig_3_3.eps } & \includegraphics[width=0.5\textwidth]{fig_3_4.eps } \end{array } $ ] solutions ( [ c3 ] ) approach the well - known@xcite exact single - soliton solutions of the static sine - gordon equation on an infinite interval:@xmath248 , ~x\equiv\frac{r\left ( \varphi-\pi\right ) } { \pi l}\in\left ( -\infty,+\infty\right ) .\ ] ] h. bluhm , n. c. koshnick , m. e. huber , and k. a. moler , phys . * 97 * , 237002 ( 2006 ) ; _ ibid_. * 98 * , 209902(e ) ( 2007 ) . it should be noted that the authors `` mimicked '' two - band superconductivity by using superconducting rings consisting of two parallel , josephson - coupled aluminium layers . however , in this particular geometrical configuration , the composite system behaves like a genuine two - band superconductor : see the discussion of the gibbs free - energy functional ( [ 2.1 ] ) . see , e.g. , l. d. landau and e. m. lifshitz , _ electrodynamic of continuous media _ ( pergamon press , oxford , 1968 ) . for reference purposes , we remark that the kinetic inductance is @xmath249 , as follows from relation ( [ 2.33 ] ) . accordingly , the expansion parameter can be expressed as @xmath250 .	in the framework of the ginzburg - landau approach , we present a self - consistent theory of specific soliton states in mesoscopic ( thin - walled ) two - band - superconducting cylinders in external parallel magnetic fields . such states arise in the presence of `` josephson - type '' interband coupling , when phase winding numbers are different for each component of the superconducting order parameter . we evaluate the gibbs free energy of the sysyem up to second - order terms in a certain dimensionless parameter @xmath0 , where @xmath1 and @xmath2are the magnetic and kinetic inductance , respectively . we derive the complete set of exact soliton solutions . these solutions are thoroughly analyzed from the viewpoint of both local and global ( thermodynamic ) stability . in particular , we show that rotational - symmetry - breaking caused by the formation of solitons gives rise to a zero - frequency rotational mode . although soliton states prove to be thermodynamically metastable , the minimal energy gap between the lowest - lying single - soliton states and thermodynamically stable zero - soliton states can be much smaller than the magnetic gibbs free energy of the latter states , provided that intraband `` penetration depths '' differ substantially and interband coupling is weak . the results of our investigation may apply to a wide class of mesoscopic doubly - connected structures exhibiting two - band superconductivity .
in a novel dynamical approach the @xmath1 vertices of a graph @xmath0 are mapped onto @xmath1 physical points located initially at equal distances from each other forming a symmetric @xmath1 simplex in @xmath2 dimensions . attractive / repulsive forces are introduced between pairs of points corresponding to connected / disconnected vertices in the original graph @xmath0 . we then let the system evolve utilizing first order aristotelian dynamics@xcite . we tune the relative strength of repulsive and attractive forces to be @xmath3 with @xmath4 the average valency i.e. average number of vertices connected to a given vertex so as to have no net average repulsion / attractions . we found , that as the system evolves various physical clusters of points tend to form . these physical clusterings reveal clusters ( or imperfect cliques ) in the graph - namely groups of vertices with a larger than average mutual connectivity . also the matrix of distances @xmath5 between the various points @xmath6 and @xmath7 is characteristic of the graph topology : points corresponding to vertices which are `` close in the graph '' namely have ( relatively ) many , short , paths connecting them will move closer together and conversely , points which are `` far in the graph '' tend to move apart . the distance matrix and clusters are important graph diagnostics . in particular the first allows us to solve easily the graph equivalence problem namely to decide if two connectivity matrices @xmath8 and @xmath9 correspond to the same topological graph and if they do to find the relabelling of vertices which makes @xmath10 and @xmath11 identical . these results are of considerable practical importance . still neither of the above problems belongs in the special class `` np complete '' problems . the latter consists of problems such as the travelling salesmen problem and the satisfiability problem for which a putative solution can be readily checked in polynomial time yet no polynomial solution method is presently known@xcite . many of these problems can be phrased in terms of graphs as the task of finding some specific graph @xmath12 inside bigger graph @xmath0 . further , all these problems which superficially seem very different are at a basic level , equally difficult : if a method of solution in polynomial time is found for one such problem then all the problems should be solvable in such time by essentially the same method . conversely if we can prove that just one npc problem necessarily require , non - polynomial time for its solution , the same holds for all of them . two of us have recently conjectured@xcite that a new variant of our approach namely of dynamically docking rigid simplexes @xmath13 and @xmath14 representing @xmath12 and @xmath0 can solve the `` g inside g '' problems . here we wish to present the first concrete application of the original , point translation or single simplex distortion algorithm ( sda ) , to an npc problems namely that of finding the largest perfect clique in @xmath0 . to most clearly illustrate the essence of the problem we consider the `` students in dorm '' example used in the general description of the clay institute prize offered for resolving @xmath15 problem@xcite . we have @xmath16 students out of which we need to select @xmath17 which can live together in a dorm , subject to a very long list of mutual exclusions . this list states that student # 1 can not be together with any one student from a specific set of say 200 other students , student student # 2 can not be together with any one from another partially overlapping set with a comparable number of students etc . how can we pick up a set of 100 students such that any one is completely compatible with the other 99 , and what is this set ? clearly this is a particular example of the general satisfiability problem where the conditions imposed are just `` two body '' exclusions . it is also a particular case of looking for a graph @xmath12 inside @xmath0 where @xmath12 is a perfect clique of vertices each of which is connected to all other members in the clique . we encode into @xmath0 with @xmath16 vertices the various mutual exclusion constraints by not connecting with edges vertices @xmath18 and @xmath19 if student # i and student # j are not compatible , and connecting by edges compatible pairs . clearly if we find within @xmath0 a clique with @xmath1 vertices it means , by our very construction , that the students to which these vertices correspond are indeed all mutually compatible . we could construct in judicious manner various smaller consistent subsets , and try piece them together . often , however a new inconsistency is revealed and we need to pursue other alternatives . while we certainly can do this in far less steps than @xmath20 the difficulty of the problem seems to grow at least exponentially with @xmath1 . in desperation we might decide to resort to the following primitive alternative and simply let the 400 students `` fight it out '' . in this all out war each student will try to push away members which are inconsistent with him and pull in those which are . this collective natural selection of the `` compatible '' - which may well be a prerelevant social phenomena - would hopefully leave us with the desired large group of mutually consistent individuals . unfortunately the outcome of such a 400 way `` somo '' fight of staying in the ring is strongly biased by the initial arbitrary placement of students in the two dimensional arena@xcite . thus we could envision a situation where an ideal group of completely compatible dorm candidates is placed in the center of a group of highly unpopular ones and is `` ejected '' together with them . in order to generate the correct large clique we need to completely unbias the starting position and avoid the severe constraints due to our existence in a physical world with limited number of dimensions . this can be done only if we go to @xmath21 dimensions and place the `` students '' which , in the inverse problem that we are really after , are metaphors for the physical points representing the @xmath22 vertices of the graph , at the vertices of a symmetric @xmath22 simplex . our search for perfect cliques uses the same physically motivated dynamical algorithm previously developed to identify via the physically bunched points clusters or `` imperfect cliques '' in a graph@xcite . we found that to adapt this algorithm for the present purpose we need only to enhance the ratio of the repulsive and attractive interactions . originally it was chosen to be : @xmath23 which could be relatively small . thus for an average valency of @xmath24 in a graph with @xmath25 vertices it is only @xmath26 . however , in order to meet the criteria of _ perfect _ cliques we clearly have to significantly enhance the strength of the repulsive interactions so as to avoid points which are connected to a fairly large number of the points in the clique but not to all of them from joining in . thus in the first round of applications we used @xmath27 we first considered a small clique of @xmath28 in a graph with @xmath29 vertices with the connectivity matrix of fig.([fig : clq - c ] ) . in addition to the clique this matrix consists of six clusters with randomly created internal connections with average valency @xmath31 . these clusters , in turn , have been randomly interconnected with a large valency @xmath32 . to simulate a real - life situation of networks with unknown structure ( topology ) we randomly permute the rows and columns of the matrix @xmath10 obtaining the reshuffled matrix @xmath33 shown in the fig.([fig : clq - b ] ) . next we apply our algorithm for clusters reconstruction using equal attractive and repulsive constant forces in @xmath34 dimensional space . the vertices of the 100-simplex were allowed to move under the influence of the forces on the 98-dimensional hyper - sphere in 99-dimensions . after a number of steps we analyzed the mutual distances between the vertices of the simplex and group neighbors which are close to each other into cliques . the new cluster - connectivity matrix is shown in fig.([fig : clq ] ) . to see how the algorithm works for the case of overlapping cliques we considered two examples . the first example includes two cliques @xmath30 and @xmath35 with a @xmath36 overlap on a `` background '' of a @xmath37 matrix with the same average @xmath32 connectivity as above . the corresponding connectivity matrixes before reshuffling is shown in fig.([fig : clq2-c ] ) . the second example involves two of cliques @xmath24 with a large overlap of @xmath39 in fig.([fig : clq3-c ] ) . our reconstruction yields one clique of @xmath24 and one of @xmath39 fig.([fig : clq3-cl ] ) . as expected we fully reconstruct the largest cliques . this is done at the expense of correspondingly reducing the size of the reconstructed part of the overlapping smaller or equal size cliques . other examples involve a @xmath17 clique in a @xmath16 graph corresponding to the `` students in dorm '' question . in addition we had an imperfect clique or cluster of @xmath41 with average valency of @xmath31 on a background of @xmath32 ( fig.([fig : clqnp - c ] ) ) . the successfully reconstructed @xmath25 clique after reshuffling is shown in fig.([fig : clqnp - cl ] ) . fig.([fig : clqnp3-cl ] ) shows the reconstructed results after reshuffling . the clique is imperfectly reconstructed . however , increasing @xmath45 from one to two improves the reconstruction as shown in fig.([fig : clqnp3-cl2 ] ) . the above results are to our mind fairly impressive . these results show that by mild tuning our original code solves in very short time and for many cases the npc problem of the largest clique . it may still fall short of solving it in all cases . as a worst case scenario we could envision a vertex ( or several such vertices ) which are connected to all the vertices in the clique save one . to avoid these vertices from joining the clique even in this case , rendering it imperfect , we need that the single repulsion due to the missing edge , overcome all the @xmath2 attractions to the rest of the points in the clique . thus the strict perfect clique worst case scenario demands @xmath46 this wildly differs from the above eq.(refx ) : for a graph @xmath0 with @xmath25 vertices @xmath47 and a clique of size @xmath48 we need a factor hundred enhancement of the ratio @xmath45 from @xmath26 to @xmath24 ! our @xmath49 based intuition would strongly suggest that this stops formation of all cliques , perfect or not , since as any given point tries move towards its `` designated '' clique it may be `` overwhelmed '' by the many repulsive forces which will prevent it from joining the clique . the configuration with the perfect clique ( and the largest perfect clique in particular ) fully formed i.e having all its vertices collapse at a point is indeed the desired final lower energy state . however there may be false local minima which trap our system just like in spin glass@xcite and protein folding problem@xcite . this is indeed most certainly the case for `` low '' dimensionalities . however with @xmath50 , as is the case here , the above intuition fails . specifically any one given `` test point '' feels just as many different forces in the directions of the other particles namely @xmath51 as there are independent directions @xmath21 to move in . ideally therefore the test particle should be able to simultaneously respond to all different @xmath51 forces , move in the direction of all the attractors and away from all the repellers and in the process further lower the energy of the system . we can adopt a local , non - orthogonal , system of coordinates where the @xmath51 axes are aligned along the unit vectors pointing from @xmath52 - the chosen point , to @xmath53 the other @xmath51 points . using our choice of constant forces@xcite we have then a net force @xmath54 which is the sum of the unit vectors along these axes with + and - signs . since these are @xmath51 linearly independent vectors the sum never vanishes @xmath55 always and no local minimum arises . there is one `` small '' correction however to the above argument . it is due to the fact that in our original algorithm we have introduced one further constraint on the motion of the points , namely that at all times on the unit circle @xmath56 . it seemed necessary in order to avoid running away to infinity of repelling vertices or collapse to the origin of attracting ones . this does however introduce an extra normal reaction force that could in fact cancel the above sum in eq.([z ] ) , and thus yields local minima . hence in the final runs we did not impose this constraint . instead we modified our code to facilitate handling the increasing distances between points at later stages of the evolution . we found that our program fully reconstructed the maximal clique@xcite . this happens regardless of the degree of the connectivity of the random background and also of the existence of large and partially overlapping slightly smaller cliques . thus for the n=100 maximal clique in an n=400 vertex graph ( i.e the students choice for dorm problem ) we added two 80x80 cliques which overlapped our 100x100 clique in two 60x60 patches which , in turn , had a 20x20 overlap and used a background with 70% connectivity fig.([fig : c3in400 ] ) . even under such seemingly unfavorable conditions we reconstructed our clique fig.([fig : cl3in400 ] ) . v. gudkov , j.e . johnson and s. nussinov , arxiv : cond - mat/0209111 ( 2002 ) . v. gudkov and s. nussinov , arxiv : cond - mat/0209112 ( 2002 ) . s. nussinov and z. nussinov , arxiv : cond - mat/0209155 ( 2002 ) . t. sudkamp , `` languages and machines : an introduction to the theory of computer science '' , addison - wesley , 1997 . clay institute , http://www.claymath.org/prizeproblems/pvsnp.htm . social pressures / encouragement via nasty / favoureble communications are not limited by dimensionality - explaining its vast power and magically efficiency . m. mezard , g. parisi and m.a . virasoro , `` spin glass theory and beyond '' , world scientific , singapore , 1987 . bryngelson and p.g . wolynes , j. phys . chem . * 93 * , p. 6902 this choice common to the present and earlier works , was originally made for simplicity . in retrospect it turns to be optimal . indeed if the forces - or potentials become too strong at short distances , then two points which attract each other , may prematurely `` seize '' and become inseparable - a phenomenon which is a source of the deleterious multiple local minima . the general convexity condition for the potentials @xmath57 which avoids this difficulty -and which our linear potential trivially satisfies- has been pointed to us by vassilios s. vassiliadis ( private communication ) . once we find a substantial portion @xmath58 of the vertices belonging in the perfect clique the construction of the rest of the clique can be easily completed as follows : we omit from our graph g all those vertices which do not connect to all the vertices in the clique . this will drastically reduce the number of vertices in the remaining relevant part of g and vastly accelerate the remaining calculations .	a novel approach to complex problems has been previously applied to graph classification and the graph equivalence problem . here we apply it to the np complete problem of finding the largest perfect clique within a graph @xmath0 .
inclined air showers are conventionally defined as those arriving at ground with zenith angles @xmath1 above @xmath2 . at large zenith angles the electromagnetic ( em ) component in air showers , mainly produced by the decay of @xmath3s , is largely absorbed in the greatly enhanced atmospheric depth the shower needs to cross before reaching ground , so that in a first approximation only the more penetrating particles such as muons survive to ground . muons are accompanied by an electromagnetic component produced mainly by muon decay in flight and muon interactions such as bremsstrahlung , pair production and nuclear interactions . a full characterisation of this so - called electromagnetic halo " is given in this work . the study of inclined showers is of great interest because their detection immediately enhances the exposure of existing air shower detectors by up to about @xmath4 with respect to that achieved with vertical showers @xmath5 , extending the field of view to sky directions otherwise inaccessible . inclined showers have been detected in the past in arrays of ground detectors such as the haverah park @xcite and agasa @xcite experiments . modern detectors such as the pierre auger observatory @xcite or the telescope array @xcite can also be used to detect showers with large zenith angles . in particular , the surface detector array ( sd ) of the pierre auger observatory is well suited to detect very inclined showers at energies above about @xmath6 ev , with high efficiency and unprecedented statistical accuracy . the energy spectrum of ultra - high energy cosmic rays ( uhecr ) with inclined showers was recently measured @xcite . the surface detector is an extended array of deep water cherenkov detectors that act like volume detectors adequate for recording particles arriving at ground at all zenith angles . as in all ground arrays , the distribution of the detector signals produced by shower particles is used to estimate shower observables such as the primary energy and to perform composition studies . as shown in previous works @xcite , the specific characteristics of inclined showers entail that their analysis requires a different approach from the standard one for vertical showers . the study of the particle densities of the electromagnetic and muonic components at ground level becomes essential in the reconstruction @xcite and analysis of events at large zenith angles . inclined showers have also been studied for years for other reasons . in the 70 s , it was suggested that showers induced by neutrinos could be identified in the background of inclined proton or nuclei induced showers by searching for deeply penetrating inclined showers @xcite which should exhibit a significant electromagnetic component at ground . in this respect the study and characterisation of the electromagnetic halo in inclined showers is also of great importance . in this paper we have used monte carlo generators simulating the air shower development to obtain the particle densities induced by the electromagnetic and muonic components of inclined air showers . in contrast to other works @xcite we have performed a comprehensive characterisation of the ratio of the electromagnetic to muonic densities as a function of distance to the shower core for different zenith angles . we have studied the dependences of this ratio on the primary energy , mass composition and hadronic interaction model assumed in the simulations . we have also studied the effect of the geomagnetic field that deviates charged particles along their paths to the detector . as a result of this study we give parameterisations of the average ratio of the em and muonic components as a function of the distance to the shower core , shower zenith angle ( @xmath1 ) and azimuth angle ( @xmath7 ) of the position at which the particle arrives at ground with respect to the incoming shower direction , as well as on the distance to the shower core in the shower plane . the effect of the intrinsic shower - to - shower fluctuations on the electromagnetic and muonic contributions to the density has also been studied . our results are relevant for the reconstruction of inclined showers in ultra - high energy cosmic ray detectors such as the pierre auger observatory and the telescope array . we have generated a library of inclined air showers induced by proton and iron primaries using the monte carlo shower propagation code aires 2.6.0 @xcite and the hadronic interaction models qgsjet01 @xcite and sibyll 2.1 @xcite . showers were generated with energies of 1 , 10 and , with zenith angle ranging from 60@xmath8 to 88@xmath8 in steps of 2@xmath8 and random azimuth angle ( unless otherwise indicated ) . we have simulated 100 showers for each combination of energy and zenith angle . showers were simulated with and without geomagnetic field assuming the location of the surface detector of the pierre auger observatory @xcite ( ground level placed at a depth of @xmath9 ) and using a curved atmosphere based on the linsley s atmospheric model @xcite . the number of particles that are produced in an air shower at eev energies and above and the computing time needed to follow all the secondaries is typically excessively large . to avoid this problem , current air shower simulation codes use a statistical sampling algorithm ( thinning algorithm ) which allows to propagate only a small representative fraction of the total number of particles @xcite . statistical weights are assigned to the sampled particles in order to compensate for the rejected ones . therefore , the output of the air shower simulation is a detailed data file with weighted entries containing the particle information at ground level . in this work , showers were simulated with a relative thinning level of @xmath10 and a weight factor for the electromagnetic ( heavy ) particles of 12 ( 0.14 ) . since the thinning process itself introduces artificial fluctuations , for the study of the intrinsic shower - to - shower fluctuations we have also simulated sets showers with a thinning level of @xmath11 . all the particles with kinetic energies above the following thresholds were tracked : 80 kev for electrons , positrons and gammas , 10 mev for muons , 60 mev for mesons and 120 mev for nucleons and nuclei . since the ground arrays differ in detection thresholds and detector response , we have performed particle energy cuts at ground level corresponding to the particular case of water cherenkov detector thresholds : 264 kev for electrons and positrons , 1.286 mev for gammas and 54.6 mev for muons . note that the kinetic energy thresholds used in the aires simulations for electromagnetic particles are all below the cherenkov thresholds , implying that we are not artificially eliminating electrons , positrons or photons that could contribute to the electromagnetic densities . concerning muons , one can see by extrapolating the energy spectrum of muons shown in fig . [ muonspectrum ] that the number of muons below 10 mev in inclined showers not accounted for in our aires simulations but still being able to produce electrons above the cherenkov threshold in water represents a small fraction of the total number of muons . to calculate the particle densities induced by the electromagnetic and muonic components of the shower , we perform a simple resampling procedure ( also called _ unthinning _ procedure ) . the particle density is calculated as the sum of the statistical weights assigned by the thinning algorithm of the particles falling in a sampling region divided by its area . the sampling area must be large enough so that a significant fraction of particles falls inside it , but at the same time it should be small enough so that particle properties are representative of their expected properties in the particular region of interest @xcite . in this work , we have chosen two different sampling regions in the plane perpendicular to the incoming shower direction ( shower plane ) . on the one hand following the previous criterion , , we have considered particles falling in concentric rings with width of 0.08 in @xmath12 in the shower plane to obtain the particle densities as a function of the distance to the shower axis ( lateral density ) . on the other hand , we have considered square cells of @xmath13 m@xmath14 in the shower plane when calculating the two dimensional distributions of the particle densities . the conventional separation between vertical and inclined showers is based on the zenith angle @xmath1 of the arrival direction of the cosmic ray particle that induces the shower . this separation stems from the different atmospheric grammage that the showers have to cross before reaching the ground , which increases approximately as @xmath15 . this fact implies that the showers arrive at ground at different stages in their evolution depending on @xmath1 . a hadronic cosmic ray typically initiates an air shower at the top of the atmosphere in the first few 100 g @xmath16 . the em component of the shower rises as the shower develops and reaches a maximum at a depth of @xmath17 for a proton shower , close to the total vertical depth of atmosphere for the altitude of the pierre auger observatory . after @xmath18 , the em component is rapidly absorbed in the atmosphere mainly due to low - energy ionization processes and the photoelectric effect . meanwhile , non - decaying muons propagate almost unattenuated to the ground , except for ionization energy losses and deflections in the geomagnetic field . consequently , a shower arriving with @xmath19 reaches the ground shortly after shower maximum and the electromagnetic component dominates at ground . however , for showers arriving with @xmath20 the atmospheric slant depth increases from @xmath21 to up to more than @xmath22 for completely horizontal showers . this results in the dominance of the muonic component at ground @xcite accompanied by a small electromagnetic component consisting mainly of the remnant of the electromagnetic shower due to cascading processes from @xmath3 decay of hadronic origin , and the electromagnetic halo due to muon decay in flight and hard muon interactions . low energy muons ( below a few gev ) typically decay along their paths to the ground generating small electromagnetic subshowers contributing to the particle densities at ground . this component mimics the muon spatial distribution and is proportional to the muon density @xcite . hard muon interactions ( pair production , bremsstrahlung and hadronic interactions ) become more and more relevant as muon energy increases . to illustrate when these processes are expected to contribute larger to the em halo , in the left panel of fig . [ muonspectrum ] we show the energy spectrum of muons at distances of 30 m and 1000 m to the shower axis for 10 eev proton - induced showers at different zenith angles . as shown here , these processes are expected to contribute to the em halo especially in highly inclined showers in which most of the muons are typically very energetic ( hundreds of gev ) , because the lower energetic muons typically decay before traveling the enlarged distances from their production height to the ground . hard muon interactions are also expected to be more frequent close to the shower core where a larger content of energetic muons is expected since energetic muons deviate less from the shower axis . a vertical shower typically exhibits a symmetric pattern of particle densities around the shower core in the shower plane . in the case of inclined showers there are several effects that produce an asymmetry on the pattern of particle densities . the most important sources of which are described below . muons in inclined showers travel along sufficiently long paths in the atmosphere to be affected by the earth s magnetic field . positive and negative muons are deviated in opposite directions with respect to the rectilinear trajectories they would follow in the absence of geomagnetic field . as a consequence the field distorts the patterns of the muonic densities in the shower plane producing elliptical or even 2-lobed patterns in very inclined showers at ground , an effect extensively studied in @xcite . the degree of distortion and the shape of the pattern depend on the strength and orientation of the magnetic field with respect to the shower axis , namely on the component of the field perpendicular to the shower axis @xmath23 , as well as on the distance traveled by the muons . this introduces a dependence of the asymmetry on the shower zenith ( @xmath1 ) and azimuth ( @xmath24 ) angles . as it was shown in @xcite the effect on the muonic distributions is only significant for @xmath25 . in fig . [ compbperp ] we show @xmath23 as a function of @xmath24 for the particular location of the southern pierre auger observatory site . for a given zenith angle , the distortion is expected to be maximal close to @xmath26 and @xmath27 and minimal around @xmath28 . @xmath29 points eastwards and is oriented counterclockwise . as an example of the effect of the geomagnetic field on the particle densities , in fig . [ muonmapwithandnob ] we show 2-dimensional maps of the muonic and electromagnetic particle densities in the shower plane in a eev proton induced shower arriving at @xmath30 and @xmath31 , with ( right ) and without ( left ) the geomagnetic field induced . the distortion of the cylindrical symmetry of the muonic component is apparent , and since the electromagnetic halo preserves the spatial distribution of muons in inclined showers , the pattern of the electromagnetic particle density exhibits a shape similar to the muonic one , although slightly blurred due to multiple coulomb scatterings suffered by the electrons before reaching ground . besides the asymmetries induced by the geomagnetic field , there is also an azimuthal asymmetry in the muonic and electromagnetic particle densities due to the combination of geometrical and shower evolution effects @xcite . as illustrated in fig . [ asymschema ] shower particles do not travel parallel to the shower axis in general and therefore they cross different amounts of atmospheric depth before reaching ground . the crossed depth depends on the azimuthal angle ( @xmath7 ) of the position on the ground at which the particle arrives with respect to the incoming shower direction . in particular , particles arrive at the ground in the _ early _ region of the shower ( the portion of the shower front that hits the ground first corresponding to @xmath32 ) more vertically than those in the _ late _ region ( corresponding to @xmath33 ) . this is essentially the basis for the geometrical effect , which depends strongly on the specific characteristics of the detectors sampling the shower front . the asymmetry induced by the geometrical effect is typically small in showers with high zenith angles . the more inclined a shower , the more energetic the muons arriving at ground are and hence the smaller the difference in the arrival angle distributions between the early and late regions of the shower . the asymmetry induced by the shower evolution can be understood as follows . particles at the same distance to the shower axis in the shower plane , but arriving with different azimuthal angles @xmath7 travel along different paths , and belong to different stages in the evolution of the shower . the importance of this effect depends on the evolution of the lateral particle distribution and on the attenuation of the total number of particles with the atmospheric depth . in fig . [ asymschema ] the detector in the early region of the shower is hit by a younger stage in the evolution of the shower than the detector in the late region . more quantitatively , for instance in a 10 eev proton shower at @xmath34 , there is a difference of @xmath35 g @xmath16 ( for notation see fig . [ asymschema ] ) between the atmosphere crossed by particles produced at the same height , traveling in straight line and hitting the ground in the late and early regions at a distance @xmath36 m to the core in the shower plane . this difference increases with the distance from the shower core . the asymmetry induced by the evolution of the shower affects more the electromagnetic component produced by @xmath37 decay than the muonic component or its associated electromagnetic halo . the reason is that this component is exponentially suppressed after the shower maximum , and small variations of the depth crossed by the shower induce large differences in the number of electromagnetic particles on the ground . however , the muonic component is less attenuated and therefore the asymmetry induced by this effect is smaller . as a consequence , shower evolution is expected to induce a much smaller asymmetry in the density in showers with zenith angles @xmath38 , because the electromagnetic component from @xmath37 decay is practically suppressed , and the electromagnetic halo is simply following the pattern of the muonic component where the asymmetry is small . this is shown in fig . [ asymdensit ] where we plot the relative difference between the electromagnetic and muonic particle densities in the early ( @xmath39 ) and late ( @xmath40 ) regions in the shower plane as obtained in our simulations . for the moment the geomagnetic field effect is neglected in this study . in the left panel of fig . [ asymdensit ] it can be seen that for showers at @xmath34 the difference in the electromagnetic density between the early and the late regions is very large even at small distances to the core ( for example a factor 1.2 at @xmath41 100 m ) , because the electromagnetic component from @xmath37 decay is still significant in the early region , and it is significantly absorbed before reaching ground in the late region . this difference increases steeply with distance @xmath42 . however , at @xmath43 the difference between the densities in the early and late regions is smaller , because the electromagnetic component from @xmath37 decay is absorbed at all @xmath7 and only the electromagnetic halo remains , except for at very far distances from the core where , besides inheriting the already quite large asymmetry due to the muonic component , there is still some remnant of the electromagnetic shower present in the early region but not in the late one . this explains why the early - late asymmetry follows essentially the behaviour of the corresponding muonic density ( right panel of fig . [ asymdensit ] ) . in the muonic case , the relative difference between the early and late densities is always small regardless of zenith angle . the electromagnetic and muonic particle densities have a characteristic behaviour with distance to the shower axis , shower zenith angle and azimuthal position with respect to the incoming shower direction . also the different contributions to the electromagnetic particle densities differ from each other as will be shown below . firstly , we have studied the em and muon number densities as a function of the distance to the shower axis ( lateral distributions ) in 10 eev proton - induced showers at different zenith angles , averaging over the azimuthal angle @xmath7 on the ground . we have also parameterised the lateral distributions of the particle densities as a function of shower zenith angle for practical applications , using 10 eev proton showers simulated with the hadronic model qgsjet01 as reference . these parameterisations are presented in appendix a. the results from these parameterisations are shown as solid lines in the top panels of fig . [ emmusignal ] compared to the simulations . the simulations are reproduced by the fits without significant deviations as shown in the bottom panels of fig . [ emmusignal ] ( @xmath44 for muons and @xmath45 for the em component ) . as shown in the left panel of fig . [ emmusignal ] , the muonic density @xmath46 decreases with @xmath1 , because muons need to travel larger distances before reaching ground arriving at larger distances to the core and being spread over a larger area . this is mainly due to the transverse momentum inherited from its parent hadron . muons of the lowest energies typically decay , decreasing further the density with theta , especially at large distances . both effects also explain the behaviour of the muon energy distribution with @xmath1 and distance to the shower axis shown in fig . [ muonspectrum ] . for example , in the right panel of this figure one can see that the mean muon energy increases with @xmath1 for a fixed @xmath42 , because only the more energetic muons survive and besides , these deviate more from the shower axis . the behaviour of the electromagnetic lateral distribution in inclined showers ( right panel of fig . [ emmusignal ] ) , can be qualitatively understood as a combination of the different behaviour of the two contributions to the em component namely , that produced by @xmath3 decay ( from hadronic origin ) and the electromagnetic halo ( from muonic origin ) . separating the two main contributions to the total em density is not possible in aires simulations because the information on the mother particle ( a @xmath3 or a muon ) producing the em subshower is lost . however we have devised a procedure to obtain in an approximate way the various contributions to the total em density . the total em density @xmath47 can be obtained as : @xmath48 where @xmath49 is the em density due to decay of @xmath3s produced in hadron and meson interactions , @xmath50 is the em density due to muon decay and @xmath51 is the em density due to muon bremsstrahlung , pair production and nuclear interactions . using aires we have performed a special subset of shower simulations in which we artificially set the muon energy threshold above which muons are explicitly followed in the simulations to a very high energy ( 10 tev ) , and at the same time we set the muon lifetime to infinity so that muons never decay . the last condition needs to be forced otherwise muons are artificially decayed in aires . the em density obtained in this way has no contribution from muons ( i.e. @xmath52 ) and can only be generated in @xmath3 decays ( produced in hadron and meson interactions ) . in fig . [ emremnant ] we show the lateral distribution of the em density as obtained in these simulations for different zenith angles . it can be seen that as the zenith angle increases from @xmath2 the em density due to @xmath3 decay is increasingly absorbed , until the zenith angle reaches @xmath53 and it practically disappears , except for a small contribution still reaching ground very near the core at @xmath54 m and far from it at @xmath55 km reaching only the early region of the shower . for @xmath38 the em component from @xmath37 decay is negligible and the em halo dominates at essentially all distances to the core . this is reflected in the fact that the em lateral distribution follows the behaviour of the muonic one . this is more apparent in fig . [ emmuratio ] where the ratio of the electromagnetic and muonic densities is shown . note that the results shown in this figure are obtained with standard aires simulations . we have also studied the behaviour of the ratio of the electromagnetic to the muonic densities : @xmath56 in fig . [ emmuratio ] , we show @xmath57 averaged over azimuth angle @xmath7 as a function of the distance to the core @xmath42 for different shower zenith angles @xmath1 and as a function of the @xmath1 for different fixed distances as obtained in the simulations , and the ratio predicted by the parameterisations proposed in appendix a. near the core , the ratio decreases with zenith angle from @xmath58 to @xmath59 due to the increasing absorption of the em component from @xmath3 decay , and then increases again with @xmath1 as can be seen in fig . [ emmuratio ] , mainly due to muon hard interactions processes that are expected to dominate near the core in very inclined showers . far from the core the lateral distribution of the ratio tends to flatten due to the dominant contribution of the em halo produced by muon decay in flight . the larger the zenith angle , the closer to shower core the ratio levels off . the slight increase of the ratio for @xmath60 and far from the core ( @xmath61 km ) is attributed to the combination of two effects , one is that the number of low energy muons decreases more rapidly at large distances because they decay before reaching the ground , and only energetic muons survive , and on the other hand the presence of the contribution to the em density due to @xmath3 decay , particularly in the early region of the shower . in addition to the dependence on zenith angle and distance to shower core , we have also studied the azimuthal asymmetry of @xmath62 . for the moment the effect of the geomagnetic field is neglected . in the left panel of fig . [ ratiomap ] , we show the 2-dimensional map of the ratio @xmath57 in the shower plane for 10 eev proton showers at @xmath34 . the shower incoming direction is from east to west in the figure . as expected , there is a clear azimuthal asymmetry at a fixed distance to the core . the contribution of the electromagnetic component is larger in the early region . however , as shown in the right panel of fig . [ ratiomap ] , at zenith angles @xmath63 the azimuthal asymmetry is less significant because only muons accompanied by the electromagnetic halo arrive at ground . since these two components approximately have the same asymmetry ( section 3.1.2 ) , the final asymmetry is practically canceled out when making their ratio . to study further the azimuthal dependence of the asymmetry we divide the shower plane in bins of width @xmath64 centered around @xmath7 , and we calculate the lateral distributions of the ratio in each bin for a fixed zenith angle : @xmath65 , and we compare these distributions to the distribution @xmath66 obtained averaging over @xmath7 . for this purpose we define the asymmetry parameter @xmath67 as @xmath68 in fig . [ exampleasyemmuratio ] , we show the lateral distribution of @xmath57 in different @xmath7 bins compared to the mean value ( left panel ) and their corresponding asymmetry parameter @xmath67 ( right panel ) for showers at @xmath34 . @xmath69 increases with distance to the core and it is larger in the early region than in the late region as expected . in addition to the dependence on position in the shower plane ( @xmath70 ) , the asymmetry parameter @xmath67 also depends on the shower zenith angle . in fig . [ amplitudeasy ] we show @xmath67 as a function of @xmath7 for a fixed distance @xmath71 m and different zenith angles . the amplitude of the asymmetry decreases as the zenith angle increases for the reasons explained above . this plot illustrates the importance of accounting for the asymmetry in the ratio when dealing with inclined showers with @xmath72 . we have parameterised the asymmetry @xmath67 as a function of the distance to the core , zenith angle and the azimuthal angle @xmath7 accounting for all the dependences above ( see appendix b ) . as an example , we show the results of the fit as solid lines in fig . [ exampleasyemmuratio ] . the simulations are reproduced by the fit without significant deviations as shown in the bottom panel of fig . [ exampleasyemmuratio ] . in the parameterisation , we have assumed that the azimuthal asymmetry @xmath73 is negligible when @xmath74 . as discussed before ( section 3.1.1 ) , the muon distributions are distorted by the presence of the geomagnetic field for zenith angles greater than @xmath75 . at these angles , the dominant contribution to the electromagnetic component at ground is due to the electromagnetic halo , which inherits the muon spatial distribution and is proportional to the muonic density , and we expect the ratio of the electromagnetic to muonic densities to maintain the symmetry in the azimuthal angle @xmath7 . for this reason , we have only studied the effect of the geomagnetic field on the ratio averaging over the azimuthal angle @xmath7 in the shower plane , for different shower zenith @xmath1 and azimuth @xmath24 angles . more precisely , for each @xmath1 we study the extreme cases where the effect is minimal , @xmath76 , and maximal , @xmath77 ( or @xmath78 ) , where @xmath24 is defined as in fig . [ compbperp ] . in the left panels of fig . [ beffratio ] we show the lateral behaviour of @xmath57 in the presence of the geomagnetic field for 10 eev proton showers . we also plot the relative difference between @xmath62 with and without geomagnetic field effect , namely : @xmath79 as can be seen in the right panels of fig . [ beffratio ] , the effect of the geomagnetic field on @xmath57 is more important near the core at all zenith angles . the reason for this is that only the highest energy muons are not significantly deflected and stay close to the core , and these are more likely to suffer hard interactions and induce an electromagnetic shower . as a consequence @xmath47 increases and at the same time @xmath80 decreases because lower energy muons are being deflected away from the core . these two effects produce an overall increase in @xmath57 . this increase is small for @xmath81 and @xmath82 m , with @xmath83 . when @xmath84 the effect starts to be important in the case of maximal deviation ( @xmath85 ) where @xmath86 for @xmath87 m , whereas for the case of minimal expected deviation ( @xmath31 ) the relative difference remains smaller at all distances . at larger angles @xmath88 , the geomagnetic field has a strong influence on @xmath57 , even when @xmath89 and the effect is expected to be minimal . it can also be seen that the larger the zenith angle , the farther from the shower core the influence of the geomagnetic field is still important for the reasons explained before . in conclusion , the effect of the geomagnetic field on the ratio of the electromagnetic to muonic number densities can be considered relevant for showers at @xmath90 . it should be noted that the rate of events at such high zenith angles detected at ground level by an array of detectors is small due to the reduced solid angle and the small size of the projection of the array onto the direction of the shower , so that this effect can be ignored for the purposes of data analysis in a first approximation . in this section we study the dependence of @xmath57 on primary energy , mass of the primary cosmic ray initiating the shower , and hadronic interaction model used to perform the simulations . these dependences are studied neglecting the effect of the geomagnetic field and averaging over azimuthal angle @xmath7 in the shower plane . the more energetic a shower , its maximum occurs deeper in the atmosphere , and therefore , the shower components are in a younger stage of evolution . the lateral distribution of the electromagnetic and muonic densities exhibits a characteristic behaviour with energy and depth of shower maximum @xcite . the lateral distribution of the charged electromagnetic component due to the cascading processes ( mainly due to @xmath3 decay ) is approximately given by , @xmath91 where the total number of particles @xmath92 depends on primary energy @xmath93 as @xmath94 . at shower maximum , @xmath95 @xcite . the lateral behaviour of the electromagnetic component @xmath96 depends on shower age ( @xmath97 ) . on the other hand , the lateral distribution of muons can be expressed as : @xmath98 where the total number of muons @xmath99 ( with @xmath100 ) increases slower with the energy than the electromagnetic component @xcite . also the lateral behaviour @xmath101 is approximately independent of the shower age @xcite . finally , as discussed before , the electromagnetic component due to muon decay in flight is proportional to the muonic density . hence , the lateral distribution of this component is expected to have the same energy dependence as the muonic one . combining all these facts , we expect the ratio @xmath102 to have a different behaviour depending on whether the electromagnetic component due to @xmath3 decay or the em halo contributes more to the total electromagnetic density . we study the energy dependence of @xmath57 performing the relative difference between the ratio at a given energy with respect to that obtained for 10 eev proton showers , @xmath103 ( see previous section ) : @xmath104 in fig . [ edep ] we show @xmath105 for @xmath106 1 eev and 100 eev showers . at @xmath107 , the energy dependence of the ratio is small @xmath108 , because the electromagnetic density being dominated by the contribution from the em halo , is roughly proportional to the muonic density , regardless of the shower energy . however for @xmath58 there is a dependence of @xmath109 on the shower energy , especially close to the shower core where the em component due to cascading processes contributes more to the em density ( see also fig . [ emremnant ] ) . in this angular range , we expect @xmath57 to behave as @xmath110 and hence to depend on shower energy . as can be seen in fig . [ edep ] , @xmath109 is larger ( smaller ) for 100 eev ( 1 eev ) showers than at 10 eev because of the larger ( smaller ) em component coming from the cascading processes in the shower which penetrates more ( less ) in the atmosphere . the dependence of @xmath105 on @xmath1 is studied in more detail in fig . [ edepvsth ] , where we plot @xmath105 in different bins of @xmath42 , as a function of @xmath1 for 1 eev and 100 eev proton showers . here , we confirm what was said before , for zenith angles above @xmath111 the em component is only due to the em halo and the ratio @xmath57 remains constant at the same level with energy , while if @xmath112 there is a dependence on energy that increases as the distance to the core decreases . at present , the chemical composition of the cosmic rays at the highest energies ( @xmath113 eev ) remains uncertain . some authors claim that cosmic rays at these energies are mainly protons @xcite , and others discuss the possibility of heavier elements such as iron nuclei @xcite . the most recent results from extensive air shower experiments do not allow to rise any firm conclusion @xcite . for this reason we have studied the dependence of the ratio @xmath114 on the mass of the primary particle initiating the shower accounting for protons and iron nuclei in our simulations . compared to a proton , an iron nucleus typically interacts higher in the atmosphere , producing a shower with a smaller depth of maximum . also applying a simple superposition model an iron nuclei produces a shower with @xmath115 more muons than a proton . as a result , an iron - initiated shower is expected to have a smaller electromagnetic density from cascading processes and a larger muonic density than a proton shower , and therefore a smaller @xmath57 as long as the em component due to cascading processes contributes more than that due to the em halo ( @xmath116 and close to the shower core ) . to demonstrate this , we have calculated the relative difference between @xmath57 in iron showers at 10 eev with respect to that obtained in 10 eev proton shower simulations @xmath117 in fig . [ massdepvsth ] we show @xmath118 as a function of zenith angle in different bins of @xmath42 . for reasons very similar to those that explain the energy dependence of @xmath57 studied before , we can explain the mass dependence . for angles larger than @xmath59 the dependence on the mass is negligible , because only the em halo is present and it is proportional to the muonic density . although the latter increases due to the larger mass of the primary , the former increases accordingly and the ratio stays roughly constant at the same level with mass . however when @xmath119 we can see a clear dependence on mass , with iron showers having smaller values of @xmath57 due to the fact that the em component is dominated by cascading processes and the iron - induced shower penetrates less in the atmosphere . at the highest energies , there is a lack of empirical knowledge about the hadronic interactions which greatly influence shower development @xcite . laboratory experiments have studied particle collisions only at centre - of - mass energies equivalent to fixed target energies of @xmath120 ev , so assumptions must still be made to perform the interactions needed in the shower simulations at the energies of interest ( @xmath121 ev ) . this fact leads to discrepancies between the different hadronic models on predictions of inelastic cross - sections and inelasticity ( multiplicity and energy of the secondaries ) ( see for instance @xcite for more details ) . these quantities determine to a large extent the longitudinal development of the air shower and , as a consequence , the number densities of the em and muonic components at ground . in this work , we compare two high energy interaction models currently used in cosmic ray physics : qgsjet01 @xcite and sibyll 2.1 @xcite . for proton primaries at 10 eev , the qgsjet model predicts showers that on average develop higher in the atmosphere and have @xmath122 more muons than showers simulated with sibyll . as a result , qgsjet predicts more muons at ground and a smaller electromagnetic component due to cascading processes . following a similar line of reasoning as in the case of the energy and mass dependence of the ratio of the em to muonic densities , @xmath57 is expected to be larger for showers simulated with sibyll at zenith angles @xmath123 , and roughly independent on the model for @xmath1 larger than @xmath111 . this behaviour can be seen in fig . [ haddepvsth ] , where we plot the relative difference @xmath124 between @xmath125 for 10 eev proton showers simulated with sibyll 2.1 with respect to the one obtained in showers simulated with qgsjet01 : @xmath126 @xmath124 is shown as a function of zenith angle in different bins of @xmath42 , to make the increasingly larger differences between sibyll and qgsjet for small distances to the shower axis more apparent , as expected from the dominance of the em component due to cascading processes near the core . summarizing this section , we have studied the effect of energy , mass composition and hadronic interaction model on the ratio of electromagnetic to muonic densities in absence of geomagnetic field effect , using the ratio obtained for 10 eev proton showers simulated with qgsjet01 as reference . combining all these dependences , we find that the extreme differences with respect to the reference ratio correspond to : * maximum : 100 eev proton showers simulated with sibyll 2.1 . * minimum : 1 eev iron showers simulated with qgsjet01 . in fig . [ largestdepvsth ] we show the relative differences between both cases and the reference ratio ( @xmath127 and @xmath128 ) as a function of @xmath1 in different bins in @xmath42 . the figure illustrates the extreme cases in the dependency of @xmath57 on energy , mass and model one should expect . in this section , we attempt to estimate the effect of physical fluctuations on the muon and electromagnetic number densities using simulations . the particle densities in individual showers are affected by different sources of fluctuation , namely : * physical ( intrinsic ) `` shower - to - shower '' fluctuations due to fluctuations in the atmospheric depth of the first interactions in the shower , fluctuations in secondary particle production , etc . , in general fluctuations in the cascading processes during the development of the shower . * `` artificial '' fluctuations due to the thinning procedure in the shower simulation . the ideal way of computing the `` shower - to - shower '' fluctuations of the electromagnetic and muonic components is to perform full ( non - thinned ) shower simulations . this is unfortunately not feasible due to the large computing time and huge disk space needed to store the information on particles at ground at the highest energies ( even for a single shower ) . one has to rely on thinned simulations in which artificial fluctuations are introduced . to overcome this problem , we have devised a simple approach based on a method given in @xcite to estimate how physical fluctuations affect the electromagnetic and muonic particle distributions when obtained from tracked particles with weights @xmath129 . here , we neglect the azimuthal asymmetry of the densities and obtain the densities in concentric rings in @xmath42 around the shower core of area @xmath130 . the particle density corresponding to @xmath131 particles with individual weights @xmath129 falling in an area @xmath130 is calculated as : @xmath132 so that the fluctuations of the particle densities stem from the fluctuations in the total particle number after accounting for the weights : @xmath133 . defining the average weight of the @xmath131 particles falling in a ring of area @xmath130 as @xmath134 , the unweighted particle number can be approximated as @xmath135 and the standard deviation of @xmath136 can be expressed as : @xmath137 assuming that the mean particle weight does not fluctuate from shower to shower as in the ideal case of an unthinned shower , and taking it to be equal to the average weight over all the particles followed explicitly in the shower simulation , we obtain : @xmath138 eq . [ sigman_shtosh ] represents an approximate way of obtaining the intrinsic physical shower - to - shower fluctuations of the muonic and electromagnetic particle densities from monte carlo simulations , which gives a good account of the physical fluctuations for thinning levels smaller than @xmath139 @xcite . in fact , in the case of an unthinned shower , in which only the intrinsic fluctuations should be present , @xmath140 , @xmath141 and eq . 8 gives @xmath142 as expected . in fig . [ signalfluct ] we show the relative physical fluctuations ( @xmath143 ) of the electromagnetic ( top ) and muon ( bottom ) densities as a function of the distance to the shower core for 10 eev proton showers at different zenith angles . these have been computed obtaining first @xmath144 and @xmath145 from the simulations and then applying eq . [ sigman_shtosh ] . the fluctuations in the muonic component are roughly independent of @xmath42 in the wide range of distances to the core plotted in the figure , and remain at the level of @xmath146 . on the other hand , the fluctuations in the electromagnetic component seem to depend on the contribution of the em halo . at @xmath34 the decrease of the relative fluctuations as the distance to the core increases could be interpreted as the em halo is becoming the dominant contribution to the em density and mimicking the behaviour of the fluctuations in the muonic component . similarly , for @xmath63 the fluctuations mimic the behaviour of those of the muonic component , they are roughly independent of @xmath42 and remain at the level of @xmath146 . we have studied the characteristics of the particle densities of the electromagnetic and muon components of inclined showers at the ground level using monte carlo simulations . we have shown that the electromagnetic component is composed of several sub - components originated in cascading and muonic processes . these different contributions to the electromagnetic component differ from each other on their behaviour with distance to the core , zenith angle and angular position ( @xmath7 ) . we have studied the ratio of the electromagnetic to muon densities ( @xmath57 ) as a function of several parameters . firstly , we have characterised the dependence of this ratio on the distance to the core and shower zenith angle ( see fig . [ emmuratio ] ) . near the core up to @xmath147 its behaviour is explained by the increasing absorption of the contribution to the em component due to @xmath3 decay and beyond @xmath148 by the dominance of the contribution to the em component due hard muon processes . far from the core ( @xmath113 km ) , the ratio is compatible with an almost constant value because the electromagnetic component due to muon decay in flight dominates the electromagnetic density at ground . then , we have studied the dependence of @xmath57 on the azimuthal position @xmath7 . for showers with @xmath72 we have found an azimuthal asymmetry , which is mostly due to the longitudinal development effect . moreover , we have studied the effect of the earth magnetic field in this ratio finding that is important for showers at @xmath90 . considering all these dependences above and neglecting the geomagnetic field effect , we propose a parameterisation of the @xmath65 that predicts the simulated data to a good precision level ( see fig . [ exampleasyemmuratio ] ) . the fits are valid for 10 eev proton showers simulated with the hadronic model qgsjet01 , and in the ranges @xmath149 $ ] and @xmath150 $ ] m. we have characterised the dependence of this ratio with the primary energy , the primary mass composition and the hadronic interaction model used in the simulations . the general result is that at zenith angles @xmath151 the ratio remains constant because only the electromagnetic halo contributes to the electromagnetic component . finally , we have estimated the effect of physical fluctuations on the electromagnetic and muon densities . the muon fluctuations remain roughly constant at the level of @xmath152 . while , the em fluctuations depend on the dominance of the electromagnetic halo on the em component , and tend to be constant @xmath146 for higher zenith angles . in table [ tabla1 ] we summarise some results for two different distances to the core and different shower zenith angles . note that in the angular region @xmath72 the presence of the remnant of the electromagnetic shower due to cascading processes from @xmath3 decay of hadronic origin entails important dependences of the ratio on the energy , mass composition and hadronic model . therefore , this region could be the most suitable one for studies about composition and hadronic interaction models by means of the electromagnetic component in the shower at ground level . however in the region @xmath153 the ratio is not sensitive to composition and hadronic model , and this angular region might be in principle more suitable to study the uhecr energy spectrum with inclined events detected by ground detectors . .summary of the dependences of the ratio of em to muonic component on the geomagnetic field , energy , primary mass and hadronic model , using protons at @xmath154 ev simulated with the qgsjet01 model and with no geomagnetic field as reference ( see eqs . [ deltab ] , [ deltae ] , [ deltam ] and [ deltah ] for the definitions of the different @xmath155 ) . the dependences are given at different distances from the shower axis 100 m and 1 km , and different zenith angles @xmath156 , @xmath148 and @xmath157 . we also give the relative physical fluctuations of the em and muonic components as obtained with eq . [ sigman_shtosh ] . the results are given in percentage @xmath158.[tabla1 ] [ cols="^,^,^,^,^,^,^,^ " , ] the authors thank gonzalo rodrguez - fernndez and our colleagues of the pierre auger collaboration for support and discussions on this topic . the helmholtz association , germany ( hhng-128 grant ) , ministerio de ciencia e innovacin ( fpa 2007 - 65114 ) , the spanish consolider - ingenio 2010 programme cpan ( csd2007 - 00042 ) , xunta de galicia ( pgidit 06 pxib 206184 pr ) and consellera de educacin ( grupos de referencia competitivos consolider xunta de galicia 2006/51 ) , and feder funds , spain , are acknowledged for providing funding . ins valio gratefully acknowledges the financial support from `` fundacin pedro barri de la maza '' ( spain ) . the authors also thank cesga ( centro de supercomputacin de galicia ) for computing resources . we give here the parameterisations performed for the lateral distributions of the electromagnetic and muonic densities at ground level projected onto the shower plane . the densities are given as a function of distance to the core @xmath42 in the shower plane and shower zenith angle @xmath1 , averaging over the azimuthal angle @xmath7 . in appendix b we also give a parameterisation of the azimuthal asymmetry . the fits are valid for 10 eev proton showers simulated with the hadronic model qgsjet01 , and in the ranges @xmath149 $ ] and @xmath150 $ ] m. we have used two different functional forms to account for the different behaviour of the lateral distributions at large distances from the core . near the core , we use a form based on the vernov functional form @xcite and for larger distances a function based on the nishimura - kamata - greisen functional form @xcite : @xmath159 with @xmath160 m. all the parameters , except @xmath161 m and @xmath162 , depend on @xmath1 . for distances @xmath163 , the parameters can be fitted with : @xmath164 @xmath165 @xmath166 for distances @xmath167 , the parameters can be fitted with : @xmath168 @xmath169 @xmath170 we have fitted the lateral distribution of the electromagnetic component to the same functional forms used for the muonic component ( see eq . [ musignals_fit ] ) and in the same radial ranges . in this case , the parameters show a different dependence on the zenith angle below and above @xmath107 , reflecting the dominance of the contribution of the em halo to the total em density . for distances @xmath163 , @xmath171 m and the remaining parameters can be fitted with : @xmath172 @xmath173 @xmath174 @xmath175 for distances @xmath167 , @xmath176 m and the remaining parameters can be fitted with : @xmath177 @xmath178 we have also parameterised the asymmetry parameter @xmath179 . the following fits are valid for 10 eev proton showers simulated with qgsjet01 , in the ranges @xmath180 $ ] , @xmath181 $ ] and @xmath150 $ ] m. note that the asymmetry is only important for @xmath182 . the parameters @xmath185 and @xmath186 can be parameterised using a cauchy - type function modified : @xmath187 \hspace{1.cm } \beta = \beta_1 \left [ \frac{1}{\pi } \left ( \frac{\beta_2}{\zeta^2 + \beta_2 ^ 2 } \right ) - \beta_3\right]\ ] ] with @xmath181 $ ] and : @xmath188 @xmath189 @xmath190 @xmath191 @xmath192 @xmath193 t. yamamoto [ pierre auger collaboration ] , proc . 30@xmath194 int . cosmic ray conference , merida , 4 ( 2007 ) , p. 335 ; j. bellido [ pierre auger collaboration ] , proc . 31@xmath194 int . cosmic ray conference , lodz ( 2009 ) , astro - ph/0906.2319 .	inclined air showers those arriving at ground with zenith angle with respect to the vertical @xmath0 are characterised by the dominance of the muonic component at ground which is accompanied by an electromagnetic halo " produced mainly by muon decay and muon interactions . by means of monte carlo simulations we give a full characterisation of the particle densities at ground in ultra - high energy inclined showers as a function of primary energy and mass composition , as well as for different hadronic models assumed in the simulations . we also investigate the effect of intrinsic shower - to - shower fluctuations in the particle densities . cosmic rays , extensive air showers , ground detector , simulation , muon component , electromagnetic component 96.50.s , 96.50.sd , 13.85.tp
the nernst effect in the normal state of high-@xmath5 cuprates has become a focus of renewed attention since the detection of an anomalous nernst signal in the normal state of underdoped cuprates @xcite . it is widely accepted that the large nernst signal is attributed to the movement of vortices which survives far above @xmath5 . on the other hand , a large negative nernst coefficient , which is comparable to the vortex signal in the superconducting state , has been reported in the charge density wave state of nbse@xmath6 @xcite . recently , it is reported that the stripe order causes the nernst signal to become large in nd or eu doped la@xmath0sr@xmath1cuo@xmath2 @xcite . thus , two mechanisms for the origin of nernst effect have been discussed up to now . the former supports the scenario that the superconducting gap opens far above @xmath5 , although the long - range phase coherence is destroyed by vortex excitation . while the latter implies that the superconductivity occurs near the quantum critical point where the competing orders , such as stripe order and/or charge order , cross in the phase diagram . here , to investigate the effects of the charge ( stripe ) ordered state on nernst signal , we controlled the strength ( stability ) of the stripe order by introducing nd in la@xmath0sr@xmath1cuo@xmath2 , and measured the nernst effect . the high - quality single crystals of la@xmath7nd@xmath8sr@xmath9cuo@xmath2 ( y=0 , 0.2 , and 0.4 ) were grown using the traveling solvent floating zone method . grown boule were carefully cut into rectangular shape confirming the crystal axis by using x - ray laue camera . all samples were carefully prepared to have the same doping level by fixing the sr content ( x=0.15 ) and the doping level was confirmed by the absolute values of the thermopower at room temperature . @xmath5 s determined by susceptibility were 37k , 23k , and 12k for y=0 , 0.2 , and 0.4 respectively . for the samples of y=0.2 and 0.4 , we observed an upturn in the resistivities around 50k and 70k , respectively , at which charge order has been observed @xcite . nernst signal @xmath10 , which is defined as the transverse electric field @xmath11 per unit temperature - gradient ( @xmath12 ) , was measured with applying three different temperature - gradients ( @xmath13 0.5 , 0.75 , 1k / mm ) . temperatures and magnetic fields were stabilized during the measurement . since the nernst signal at high temperature is linear to magnetic field , we can define the nernst coefficient @xmath14 as @xmath15 , where @xmath16 . figure 1 shows the temperature dependence of @xmath14 in high temperature . to clarify the doping dependence , @xmath14 of nd - free sample with @xmath17=0.125 is also plotted . note that the full scale of the vertical axis is as small as 20 nv / kt , which assures that the resolution of the measurement is better than 5 nv / kt . as seen in the figure , @xmath14 s of x=0.15 are completely coincide irrespective of nd content . the magnitude of @xmath14 is almost consistent with the previous report @xcite . these results suggest that the absolute value of @xmath14 at high temperatures is determined only by the doping level @xmath17 . more interestingly , the onset temperature @xmath3 , below which @xmath14 starts to deviate from high - temperature @xmath18-linear behavior ( normal - term nernst coefficient ) , does not change with nd - doping , even though the superconducting transition temperature @xmath5 is drastically reduced . of la@xmath19nd@xmath8sr@xmath20cuo@xmath2 in high temperature.,width=257 ] to investigate the effect of the stripe order , we show the temperature dependence of @xmath10 at 9 t in fig . the magnitude of @xmath10 drastically decreases with nd - doping below the temperature @xmath21 , where the magnetic field dependence of the nernst coefficient was observed ( @xmath21 was determined as ref.[3 ] ) . this result indicates that the fluctuation of superconductivity is suppressed by the stabilization of the stripe order due to nd - doping below @xmath21 . as seen in la@xmath22eu@xmath23sr@xmath1cuo@xmath2 of ref.[3 ] , we observed the evolution of two - peak structure in the sample of x=0.4 . furthermore , an upturn of the nernst signal has been clearly observed below the charge order temperature @xmath4 ( see the inset of fig.2 ) . this is crucial evidence that the static stripe order enhances the nernst signal . based on the above results , we can define three characteristic temperatures : @xmath24 . with decreasing temperature , the following scenario is expected at this stage : on the assumption that the enhancement of the nernst signal is attributed to the effect of the stripe order at high temperatures , the fluctuation of the stripe order develops below @xmath3 , and then the static stripe order enhances the nernst signal further at @xmath4 . with further decrease of temperature , a huge nernst signal induced by the vortex flow is expected below @xmath21 . at 9 t for la@xmath7nd@xmath8sr@xmath9cuo@xmath2 . @xmath21 was determined as ref.[3 ] . inset : nernst signal e@xmath8 at 9 t for y=0.4 . to confirm reproducibility , we measured two different sample.,width=238 ] in order to investigate the effect of the charge ( stripe ) order on nernst signal , we measured the temperature and magnetic - field dependences of the nernst effect in nd - doped la@xmath0sr@xmath1cuo@xmath2 single crystals with controlling the strength ( stability ) of the charge order . we found that the magnitude of nernst signal drastically decrease with nd doping below @xmath21 , although @xmath3 does not change conspicuously , and that the nernst signal is also enhanced at the charge ordered transition temperature @xmath25 . these results suggest that the stripe order as well as its fluctuation enhances the nernst signal at high temperatures , whereas the stabilized stripe order tends to suppress the superconducting fluctuation and hence to reduce the nernst signal at low temperatures . z. a. xu et al . , nature , * 406 * ( 2000 ) 486 . r. bel et al . * 91 * ( 2003 ) 066602 . olivier cyr - choini@xmath26re et al . , nature , * 458 * ( 2009 ) 743 . n. ichikawa et al . , phys . * 85 * ( 2000 ) 1738 .	we have precisely measured the nernst effect in nd - doped la@xmath0sr@xmath1cuo@xmath2 single crystals with controlling the strength ( stability ) of the stripe order . we found that the onset temperature @xmath3 , where the nernst signal starts increasing , does not change conspicuously in spite of nd - doping . at low temperatures , on the other hand , the absolute value of the nernst signal is strongly suppressed in accordance with the strength of the stripe order . these results imply that the fluctuation of ( charge ) stripe order enhances the nernst signal below @xmath3 at high temperatures , and then the stripe order enhanced by nd - doping suppresses the superconducting fluctuation to reduce the nernst signal at low temperatures . we also observed an increase of the nernst signal below the charge order temperature @xmath4 which is observed in diffraction measurement . nernst effect , vortex flow , stripe order 74.25.fy , 74.72.dn , 74.62.dh
the latest trace and soho / eit observations have brought the solar corona into sharp focus . the observations reveal : 1 ) the structures that constitute the solar corona are in constant motion ; they are full of fast moving gases , and are heated primarily at their footpoints ( base ) very close to the solar surface . the heating occurs in few minutes in the first ten to twenty thousand kilometers above the surface , i.e , in a rather small fraction of the bright part of the anchored structure . in direct contradiction to the predictions of some theories , the heating is neither uniform ( throughout the loops ) nor does it happen preferentially near the top . a direct quote sums up the situation aptly : moreover , not only heat is deposited low down , but the gas is often actually thrust upward very rapidly . it does not merely ` evaporate ' into the coronal structures , it is often actually thrown up there . exactly how that happens is still a puzzle @xcite , 2 ) the loops are composed of clusters of filamentary structures which are not , as believed before , static bodies supported by interior gas pressure and heated along their lengths @xcite . they fill and drain so quickly that the gas in them must be moving nearly ballistically ( see latest trace news and e.g. @xcite ) along the substructures , rather than being quiescently heated . from a detailed study of the loops with different characteristic parameters one concludes that the heating process is quite non uniform @xcite . transient brightenings , with their associated flows of cool and hot material , are also a very common phenomenon in the trace movies . these relatively fast ( violent ) happenings vary from small events in the quiet sun to major flares in active regions ; brightenings which are more than @xmath5 km apart often occur within the same exposure that typically lasts for 10 to 30s @xcite . this kind of a coincidence in the events at distant locations is suggestive of fast particle beams propagation along separate magnetic loops which come together at the flaring site . the flaring sites are generally assumed to be reconnection sites although observations have not establish a causal connection : direct evidence for reconnection in flares is difficult to find , despite the fact that it is thought to be the primary process behind flares it is remarkable that often the post flare loop systems begin to glow at the trace euv wavelengths without substantial distortion : reconnection that probably took place appears to be ( largely ) completed by the time the loops are detected . these observations pose a new challenge for the theories of quiescent as well as not so quiescent coronal structures and events . in this paper we examine the conjecture that the formation and primary heating of the coronal structures as well as the more violent events ( possibly flares , erupting prominences and coronal mass ejections ( cmes ) ) are the expressions of different aspects of the same general global dynamics that operates in a given coronal region @xcite . it is stipulated that the coronal structures are created from the evolution and re organization of a relatively cold plasma flow emerging from the subcoronal region and interacting with the ambient solar magnetic field . the plasma flows , the source of both the particles and energy ( part of which is converted to heat ) , in their interaction with the magnetic field , also become dynamic determinants of a wide variety of plasma states ; it is likely that this interaction may be the cause of the immense diversity of the observed coronal structures @xcite . preliminary results from this magneto fluid approach reproduce several of the salient observational features of the typical loops : the structure creation and primary heating are simultaneous the heating takes place ( by the viscous dissipation of the flow kinetic energy ) in a few minutes , is quite non - uniform , and the base of the hot structure is hotter than the rest . we plan to extend the scope of the magneto - fluid theory beyond the creation of the semi - quiescent coronal structures by seeking answers to the following : a ) can the basic framework of this model predict the possibility of , and the pathways for the occurrence of sudden , eruptive , and catstrophic events ( such as flares , eruptive prominences , cmes ) in the solar atmosphere , b ) does the eventual fate , possibly catstrophic reorganization , of a given structure lie in the very conditions of its birth , c ) is it possible to identify the range and relative values of identifiable physical quantities that make a given structure prone to eruption ( flaring ) , d ) will an eruption be the result of the conversion of excess magnetic energy into heat and bulk plasma motion as is generally believed to happen in the solar atmosphere @xcite-@xcite ? we begin by identifying the quasi equilibrium state of a typical coronal sturcture with a slowly changing double beltrami ( db ) state ( one of the simplest , non trivial magnetofluid equilibrium ) . the slow changes may be due to changes in the sun which affect the local magnetic fields , the interaction of various nearby structures , or disturbances in the solar atmosphere . the parameter change is assumed to be sufficiently slow that , at each stage , the system can find its local db equilibrium ( adiabatic evolution ) . the slow evolution must conserve the dynamical invariants : the helicity @xmath6 , the generalized helicity @xmath7 , and the total ( magnetic plus the fluid ) energy @xmath8 . the problem of predicting sudden events ( e.g. catastrophic eruption ) then reduces to finding the range , if any , in which the slowly evolving structure may suffer a loss of equilibrium . the signature of the loss of equilibrium is quite easy to identify for the db states . the transition may occur in one of the following two ways : 1 ) when the roots of the quadratic equation , determining the length scales for the field variation , go from being real to complex ( implying change in the topology of the magnetic and the velocity fields boundary separating the paramagnetic from the diamagnetic ) , or 2 ) the amplitude of either of the two states ceases to be real . for our current problem , the sudden change is likely to follow the second route . by analysing a simple analytically tractable model , we find affirmative answers to all the four questions we posed . we show that the invariants @xmath9 , and @xmath8 , which label and alongwith the initial and bounbary conditions determine the original state , hold the key to the eventual fate of a structure . if for a given equilibrium sequence , the total energy @xmath8 is larger than some critical value ( given in terms of @xmath10 , and @xmath11 ) , the catastrophic loss of equilibrium could certainly occur . the trigger for the equilibrium loss could come , for instance , from nearby structures getting close to each other with an increase in their interaction energy . the catastrophy pushes a db state to relax to a minimum energy single beltrami field . for coronal structures , the transition transfers almost all the short scale magnetic energy to the flow energy . within the framework of our approach , there are two distinct scenerios for eruptive events : a ) when a `` slowly '' evolving structure finds itself in a state of no equilibrium , and b ) when the process of creating a long lived hot structure is prematurally aborted ; the flow shrinks / distorts the structure which suddenly shines and/or releases energy or ejects particles . the latter mechanism requires a detailed time dependent treatment and is not the subject matter of this paper . the following semi equilibrium , collisionless magneto fluid treatment pertains only to the former case . a given structure is supposed to corresspond to the equilibrium solutions of the two fluid system @xmath12 written in terms of a pair of generalized vorticities @xmath13 , and effective flows @xmath14 , with the following normalizations : the magnetic field @xmath15 to an appropriate measure of the magnetic field @xmath16 , the fluid velocity @xmath17 to the corresponding alfvn speed , and the distances to the collisiolless ion skin depth @xmath18 . the simplest and perhaps the most fundamental equilibrium solution to ( [ vdynamics1 ] ) is given by the `` beltrami conditions '' , which imply the alignment of the vorticities and the corresponding flows ( @xmath19//@xmath20 ) , @xmath21 with @xmath22 . we have used constant density assumption for simplicity - extension to varying density is straightforward @xcite . equations ( [ eq : d - beltrami1 ] ) and ( [ eq : d - beltrami2 ] ) combine to yield : @xmath23 where @xmath24 , @xmath25 = `` curl '' , and @xmath26 . \label{eq : lambda_pm}\ ] ] for sub alfvenic flows ( the flows we generally encounter in the solar atmosphere ) , the length scales ( @xmath27 ) are quite disparate . we assume @xmath28 without loss of generality . the general solution to the `` double beltrami equations '' ( [ eq : d - beltrami3 ] ) is a linear combination of the single beltrami fields @xmath29 satisfying @xmath30 . thus , for arbitrary constants @xmath31 , the sum @xmath32 solves ( [ eq : d - beltrami3 ] ) , and the corresponding flow is given by @xmath33 . the db field encompasses a wide class of steady states of mathematical physics from the force free paramagnetic field to the fully diamagnetic field . the beltrami conditions also demand `` generalized bernoulli conditions '' which allow pressure confinement when an appropriate flow is driven @xcite ( and references therein ) . the db field is characterized by four parameters : @xmath34 ( eigenvalues ) , and @xmath35 ( amplitudes ) . the three invariants @xcite : the helicity @xmath6 , the generalized helicity @xmath7 , @xmath36 ( @xmath37 is the vector potential ) , and the total energy @xmath38 will provide three algebraic relations connecting them @xcite . to predict the possibility of an eruptive event , interpretted as the termination of an equilibrium sequence ( for solar flares , this kind of an approach , albeit in different contexts , has been followed in numerous investigations , ( see e.g. @xcite-@xcite and references therein ) , we analytically investigate this system using the macro scale ( @xmath39 ) of the closed strcuture as a control parameter . this choice is physically sensible and is motivated by observations because in the process of structure structure interactions , initial shapes do undergo deformations / distortions with rates strongly dependent on the initial and boundary conditions . for simplicity we explicitly work out the system in a cartesian cube of length @xmath40 . we take @xmath41 to be the simple 2-d beltrami abc field @xcite , @xmath42 with the normalization @xmath43 . for real @xmath44 , ( [ eq : abc-1 ] ) represents an arcade magnetic field structure resembling interacting coronal loops [ in fig . 1 ] . assuming @xmath45 ( @xmath46 are integers ) , @xmath47 satisfy the following relations : @xmath48 , where @xmath49 . the invariants can now be readily evaluated and the results can be displayed in several equivalent forms . we find the following three equations to be the most convenient for further analysis ( @xmath50 , @xmath51 ) : @xmath52 \nonumber \\ & & \quad \quad -\lambda _ { + } \lambda _ { -}h_{1},\label{eq : h2 - 1}\\ c_{+}^{2 } & = & d^{-1}\left\ { e-\left [ 1 + ( \lambda _ { -}+\tilde{a})^{2}\right ] \lambda _ { -}h_{1}\right\ } \lambda _ { + } , \label{eq : c1 - 2 } \\ c_{-}^{2 } & = & - d^{-1}\left\ { e-\left [ 1 + ( \lambda _ { + } + \tilde{a})^{2}\right ] \lambda _ { + } h_{1}\right\ } \lambda _ { -},\label{eq : c2 - 2}\end{aligned}\ ] ] where we have removed the common factor @xmath53 , and @xmath54 \lambda _ { + } -\left [ 1+(\lambda _ { -}+\tilde{a } ) ^{2}\right ] \lambda _ { - } \nonumber \\ & = & ( \lambda _ { + } -\lambda _ { - } ) b(b+\tilde{a } ) . \label{eq : assume - ab1}\end{aligned}\ ] ] for given @xmath6 , @xmath8 , @xmath55 ( @xmath7 ) and @xmath56 ( control parameter ) , we can solve the preceding system to determine the physical quantities @xmath57 , and @xmath58 which must all remain real for an equilibrium . before we give an analytic derivation for the bifurcation conditions ( leading to loss of equilibrium ) , we display in fig.2 the plots of @xmath57 and @xmath59 as functions of @xmath56 for two distinct sets for the values of the invariants : we choose @xmath60 , @xmath61 , @xmath62 for fig.2(a ) , and @xmath60 , @xmath61 , @xmath63 for fig.2(b ) ( dashed lines correspond to the region of imaginary @xmath64 ) . we find that the behavior of the solution changes drastically with @xmath8 . for the parameters of fig.2(a ) , @xmath57 and @xmath59 remain real and change continuously with varying @xmath56 implying that as the macroscopic scale of the structure ( @xmath65 ) changes continuously , the equilibrium expressed by ( [ eq : abc-1 ] ) persists there is no catastrophic or qualitative change . for fig.2(b ) with @xmath8 changing from 0.4 to 1.3 ( with same @xmath66 ) we arrive at a fundamentally different situation ; when @xmath56 exceeds a critical value @xmath67 , i.e , the macro scale becomes smaller than a critical size , the physical determinants of the equilibrium cease to be real ; the sequence of equilibria is terminated . the condition for catasrophe turns out to be a constraint involving @xmath68 and @xmath8 , which will allow the vanishing of @xmath69 for positive @xmath70 , and real @xmath71 . it is straightforward to show that the system has a critical point if @xmath72 and the critical @xmath73 is determined by a simultaneous solution of ( 11 ) and @xmath74 \lambda _ { + } h_{1}=0 $ ] giving : @xmath75 thus , for @xmath76 ( determined by helicities @xmath6 and @xmath7 ) , when the macroscopic size of a structure shrinks below a critical value , it can go through a severe reorganization . at the critical point , an expected but most remarkable transition occurs . using the value of @xmath77 , we find from equation([eq : c2 - 2 ] ) that the coefficient @xmath64 , which measures the strength of the short scale fields , identically vanishes , and the equilibrium changes from double beltrami to a single beltrami state defined by @xmath78 , i.e. , @xmath79 ( @xmath80 ) with @xmath17 parallel to @xmath15 . the transition leads to a magnetically more relaxed state with the magnetic energy reaching its minimum with appropriate gain in the flow kinetic energy ( see fig.[fig : solar ] ) . by modelling quasi equilibrium , slowly evolving coronal structures as a sequence of double beltrami magnetofiuid states in which the magnetic and the velocity field are self consistently coupled , we have shown the possibility of , and derived the conditions for catastrophic changes leading to a fundamental transformation of the initial state . the critical condition comes out as an inequality involving three invariants of the collisionless magnetofluid dynamics . when the total energy exceeds a critical energy the db equilibrium suddenly relaxes to a single beltrami state corresponding to the large macroscopic size . all of the short scale magnetic energy is lost having been transformed to the flow energy and partly to heat via the viscous dissipation of the flow energy . this general mechanism in which the flows ( and their interactions with the magnetic field ) play an essential role could certainly help in advancing our understanding of a variety of sudden ( violent ) events in the solar atmosphere like the flares , the erupting prominences , and the coronal mass ejctions . the connection of flows with eruptive events is rather direct : it depends on their ability to deform ( in specific cases distort ) the ambient magnetic field lines to temporarily stretch ( shrink , destroy ) the closed field lines so that the flow can escape the local region with a considerable increase in kinetic energy in the form of heat / bulk motion . for smm this study was supported by us department of energy contract no.de-fg03-96er-54366 . the work of zy was supported by grant in aid for scientific research from the japanese ministry of education , science and culture no.09308011 . work of nls was partially supported by the joint intas georgian grant no.52 . authors thank abdus salam international centre for theoretical physics , trieste , italy , where this work was started . c. j. schrijver , a. m. title , t. e. berger , l. fletcher , n.e . hurlburt , r.w . nightingale , r.a . shine , t.d . tarbell , j. wolfson , l. golub , j.a . bookbinder , e.e . deluca , r.a . mcmullen , h.p . warren , c.c . kankelborg , b.n . handy and b. depontieu , solar phys . * 187 * , 261 ( 1999 ) . r. rosner , w.h . tucker , & g.s . vaiana , astrophys . j. * 220 * , 643(1978 ) . aschwanden , r.w . nightingale & r.w . j. * 541 * , 1059 ( 2000 ) . mahajan , r. miklaszewski , k.i . nikolskaya & n.l . shatashvili . phys . plasmas , * 8 * , no.4 , 1340 ( 2001 ) . s. m. mahajan & z. yoshida , phys . lett . * 81 * , 4863 ( 1998 ) . z. yoshida & s. m. mahajan , j. math . phys . * 40 * , 5080 ( 1999 ) . t. sakurai , solar phys . * 121 * , 347 ( 1989 ) . choe & c.z . j. * 441 * , 449(2000 ) . j. * 471 * , 489 ( 1996 ) . j.birn , j. t. gosling , m. hesse , t.g . forbes & e.r . j. * 541 * , 1078 ( 2000 ) . chen & k. shibata . j. * 545 * , 524 ( 2000 ) . c. b. roald , p.a . sturrok & r. wolfson . astrophys . j. * 538 * , 960 ( 2000 ) . l. c. steinhauer & a. ishida , phys . lett . * 79 * , 3423 , ( 1997 ) . z. yoshida , s.m . mahajan , s. ohsaki , m. iqbal & n. shatashvili . phys . plasmas , * 8 * , no.5 , ( 2001 ) . t. g. forbes & p. a. isenberg , astrophys . j. * 373 * , 294 ( 1991 ) . k. kusano , y. suzuki & k. nishikawa . j. * 441 * , 942 ( 1995 ) . arnold & b.a . khesin . `` topological methods in hydrodynamics '' . ( springer verlag new york berlin heidelberg 1998 ) , p.72 .	by modelling the coronal structures by `` slowly '' evolving double beltrami two fluid equilibria ( created by the interaction of the magnetic and velocity fields ) , the conditions for catastrophic transformations of the original state are derived . it is shown that , at the transition , much of the magnetic energy of the original state is converted to the the flow kinetic energy . _ subject headings : _ sun : flares sun : corona - sun : magnetic fields ' '' '' + @xmath0 0.1 cm plasma physics department , tbilisi + 1.6 cm state university , tbilisi 380028 , georgia + @xmath1 e mail : [email protected] + @xmath2 e mail : [email protected] + 1.6 cm [email protected] + @xmath3 e mail : [email protected] + @xmath4 e mail : [email protected] +
extended air showers ( eas ) are originated by highly energetic cosmic rays , which interact with air atoms in the earth s atmosphere , producing elementary hadrons and nuclear fragments . neutral pions immediately decay in two photons , which in turn interact with other particles , generating electromagnetic cascades , while charged pions and kaons , as well as other hadrons , interact and/or decay , depending on their energy and air density , leading to the hadronic component of the shower . thus , to understand a process as complex as an air shower , models and codes capable of describing the evolution of both the electromagnetic and the hadronic component are needed . monte carlo codes have been developed in the last few years for this purpose , such as the corsika simulation package @xcite , allowing to choose among different models and model combinations for the description of eas formation and propagation . in this paper , the fluka monte carlo code @xcite is applied to the study of eas induced by cosmic rays with primary energies up to 1000 tev . these results rely entirely on the use of the fluka code for transport , interactions and decays of leptons and @xmath0s at all energies and for hadron - nucleus interactions at energies below 20 tev , while the dpmjet code @xcite is used for all nucleus - nucleus interactions and for hadron - nucleus interactions at the highest energies . our simulations are completely independent with respect to corsika , which contains only a partial version of fluka , for use at low energies only ( e @xmath1 100200 gev ) . the energy range spanned in this work is of interest for cosmic ray experiments aiming at the determination of the primary spectra and composition for energies up to the knee region , such as the atic , the runjob and the kascade experiments . it is worth mentioning that fluka has not been specifically designed for the study of cosmic ray physics , but can be applied also in this field . at energies @xmath2 30 tev , interesting results concerning the prediction of inclusive lepton fluxes have already been obtained . in particular , the first 3-dimensional calculation of the atmospheric neutrino flux due to cosmic rays @xcite was made with fluka , and , more recently , data concerning muons detected by the balloon - borne bess spectrometer at various depths and by the l3+c spectrometer located at cern have been quite succesfully reproduced @xcite . in this section some aspects of the fluka models relevant for cosmic ray description are briefly discussed , together with the validation of the models by means of data collected at accelerators . actually , even the most recent accelerator data allow to test theoretical monte carlo models only for some specific projectile beam - target combinations at energies well below the maximum primary energies , while , for the simulation of eas processes , one has to extrapolate the models to many more combinations and to the highest energies , on the basis of theoretical considerations . double differential production cross section for 12.9 gev / c protons on aluminum for different angular ranges , compared with harp experimental data @xcite . [ fig : harp ] , title="fig:",scaledwidth=23.0% ] double differential production cross section for 12.9 gev / c protons on aluminum for different angular ranges , compared with harp experimental data @xcite . [ fig : harp ] , title="fig:",scaledwidth=23.0% ] double differential production cross section for 12.9 gev / c protons on aluminum for different angular ranges , compared with harp experimental data @xcite . [ fig : harp ] , title="fig:",scaledwidth=23.0% ] double differential production cross section for 12.9 gev / c protons on aluminum for different angular ranges , compared with harp experimental data @xcite . [ fig : harp ] , title="fig:",scaledwidth=23.0% ] and @xmath3 production for proton interactions on carbon at 158 gev / c , as measured by na49 @xcite ( symbols ) and predicted by fluka ( histograms ) . linear scale on the left , logarithmic scale on the right . [ na49fig],title="fig:",scaledwidth=23.0% ] and @xmath3 production for proton interactions on carbon at 158 gev / c , as measured by na49 @xcite ( symbols ) and predicted by fluka ( histograms ) . linear scale on the left , logarithmic scale on the right . [ na49fig],title="fig:",scaledwidth=23.0% ] the modern fluka @xcite is a general purpose monte carlo transport and interaction code developed since 1988 mostly by infn and cern researchers , and used in many fields of physics , both for fundamental research and for applications . the history of the code dates back to the sixties when j. ranft developed and applied the very first version to accelerator shielding . the most common fluka applications are accelerator physics , calorimetry , shielding design , radioprotection and dosimetry , while new ones , such as hadrontherapy , are promisingly growing . since a few years , fluka is also applied to cosmic ray physics , mainly with focus on low energy cosmic rays , as already mentioned in the introduction , and on their effects for dosimetry @xcite and space physics @xcite . all these applications require both precise physics models , and packages to build geometries , to simulate targets and experimental setups . both topics are the subject of constant development and improvement in fluka . as far as the geometry is concerned , a combinatorial geometry package is included into the code , allowing to create complex geometries , which can be visualized by means of apposite graphical tools . this allows to simulate 3-d particle propagation , and , as far as cosmic rays are concerned , to study showers characterized by whichever inclination angle and direction with respect to the earth , and to take into account in a precise way the effect of different atmospheric compositions and magnetic field configurations . the same code can be used for the simulation of cosmic ray detectors , including surrounding materials . the fluka physical models are based as far as possible on theoretical microscopic models , with the advantage , with respect to parameterized inclusive models , of preserving correlations , and of being predictive in the regions where experimental data are not available . the model parameters are fixed once , internally to the code , for all projectile - target combinations and energies , and can not be modified by the user . the theoretical models are continuously benchmarked against newly available experimental data . the code is under continuous development and made available on the website _ http://www.fluka.org_. the development and maintenance are performed under a infn cern agreement . for historical reasons , fluka is best known for its hadron event generators , but since more than 17 years fluka can handle with similar or better accuracy electromagnetic effects @xcite . briefly , the energy range covered by this sector of fluka is very wide : the program can transport photons and electrons over about 12 energy decades , from 1 pev down to 1 kev . part is fully coupled with the hadron sector , including the low energy ( i.e. @xmath2 20 mev ) neutrons . the simulation of the electromagnetic cascade in fluka is very accurate , including the landau - pomeranchuk - migdal effect and a special treatment of the tip of the bremsstrahlung spectrum . electron pairs and bremsstrahlung are sampled from the proper double differential energy - angular distributions improving the common practice of using average angles . in a similar way , the three - dimensional shape of the hadronic cascades is reproduced in detail by a rigorous sampling of correlated energy and angles in decay , scattering , and multiple coulomb scattering . bremsstrahlung and direct pair production by muons are modelled according to state - of - the - art theoretical description and have been checked against experimental data @xcite . muon photonuclear interactions are also modelled . a basic description of hadronic interactions in fluka can be found in @xcite . hadron - nucleon interactions at energies below a few gev are simulated in fluka by the isobar model , through resonance production and decay , and by taking into account elastic , charge and strangeness exchange . elementary hadron - hadron collisions at energies above a few gev are described thanks to an implementation of the dual parton model ( dpm ) @xcite , coupled to a hadronization scheme . the dual parton model allows for the description of soft collision processes ( i.e. processes characterized by low @xmath4 , @xmath5 , where @xmath6 is a momentum scale of the order of a few gev ) , that can not be described by means of pqcd , due to fast rise of @xmath7 with decreasing momentum . dpm incorporates analiticity , duality and unitarity . incoming hadrons are described by strings , which interact by exchanging closed loop excitations , called pomerons . at low energies , the single - pomeron exchange is the dominant contribution , while at laboratory energies @xmath8 1 tev higher - order contributions , corresponding to multi - pomeron exchange , become increasingly important . corresponding to each pomeron cut , built according to the recipes provided by the abramoviskii - gribov - kancheli cutting rules , colorless chains are built , extending from the projectile valence and sea quarks and antiquarks to the target ones . in particular , the valence quarks in each baryon are treated as a quark - diquark pair , so that , in the simplest case of the single - pomeron exchange amplitude for a baryon - baryon interaction , two chains are built , each extending from a valence quark to a valence diquark . sea quarks ( and the corresponding antiquarks ) are additionally involved as end - points of chains in multi - pomeron exchange amplitudes . besides determining the number of pomeron cuts and forming the chains , the dpm allows to assign an energy and momentum to each chain , according to the momentum distribution functions of the partons in the hadrons . in the asymptotic regime parton masses are neglected , however , at lower energies , finite mass effects have to be included both in the chain building process , by suppressing the chains with invariant masses below the observed baryon and meson mass values and reassigning energy and momentum to other chains to ensure energy / momentum conservation , and in the hadronization procedure . hadronization is not included in dpm , but is performed by different models properly coupled to dpm . at present , the hadronization scheme implemented in fluka is based on an advanced version @xcite of the model by @xcite , with special emphasis to provide accuracy in the low - energy description of hadron formation ( down to a few gev in the lab ) . hadron - hadron collisions are the main building blocks of hadron - nucleus collisions . multiple collisions of each hadron with the nuclear constituents are taken into account by means of the glauber - gribov mechanism . particular efforts are devoted to the study of nuclear effects on hadron propagation . in the last fluka version the propagation in the nucleus of the hadrons resulting from elementary multiple collisions can be simulated with improved accuracy with respect to the past , by means of an improved generalized intranuclear cascade ( ginc ) model , followed by pre - equilibrium and coalescence and by de - excitation of the remaining nuclear fragments . all these models are included in the fluka submodule peanut , which has been refined for several years to describe with increasing accuracy many low - energy processes involving nucleons , and has been further extended in the last few months to higher energy events ( @xmath9 @xmath10 a few gev ) . more detail about peanut and the issues concerning its extension can be found in @xcite . peanut is used also to simulate @xmath0 , @xmath11 and stopping @xmath12 interactions , and nucleon decay . photon interactions with hadrons and muon virtual photon nuclear interactions are simulated by means of the vector dominance model . hadron free paths inside a nucleus can be different from those computed on the basis of free hadron - nucleon scattering , due to pauli blocking , coherence length , formation and multibody processes , all implemented in fluka @xcite . in particular , coherence length affects final states from elastic and charge - exchange scatterings : interaction products can not be localized better than the position uncertainty related to the 4-momentum transferred in the collision , according to the @xmath13 relationship . thus , reinteractions occurring at distances shorter than the coherence length undergo interference and can not be treated as independent . on the other hand , formation zone affects the reinteraction probability of secondaries emerging from high energy interactions . it is supported by experimental results , which indicate that high - energy secondary particle reinteractions inside nuclei are strongly suppressed . taking into account that a typical time for strong interactions is @xmath14 1 fm / c , a particle emerging from an interaction requires some time to materialize , according to the uncertainty principle . examples of performance of the fluka hadronic model , in reproducing experimental data recently obtained by the harp and na49 collaborations are shown in fig . [ fig : harp ] and fig . [ na49fig ] . both hadronic physics at energies above 20 tev and heavy - ion collisions above 5 gev have been simulated in this work thanks to the interface of fluka with the dpmjet code . dpmjet - ii.53 and dpmjet - iii libraries can be both used within fluka . dpmjet is also available in corsika as a possible choice among multiple high - energy event generators , and is widely used in the cosmic ray physics community . dpmjet has been extensively tested against data from the sps collider , the tevatron and rhic . soft physics is described in dpmjet @xcite thanks to the dpm , already introduced in the previous section . furthermore , hard physics processes are implemented , based on leading order pqcd . soft and hard processes contribute together to the inelastic cross - section according to the eikonal approximation , which provides a unitarization scheme . the eikonal function is given by the sum of the soft and hard components , also including terms from multi - pomeron exchange . one of the main differences between the fluka hadronic physics models and dpmjet is related to the hadronization process . in dpmjet this process is performed by means of the jetset chain fragmentation code included in pythia @xcite , with some modifications . in particular , while the pythia code was extensively tested for processes involving a significant hard component , the transverse momentum distribution in soft chain decay has been modified in the jetset version included in dpmjet to better describe soft chain hadronization . this way , the transverse momentum distributions of k and @xmath15 predicted by dpmjet - iii show an improved agreement with experimental data . there are other differences in the description of final state interactions : dpmjet has its own ginc code , simplified with respect to the fluka one , and does not take into account coalescence and pre - equilibrium emission . nuclear de - excitation is performed here by means of the same de - excitation models @xcite used in fluka . further refinements of dpmjet and of its interface to fluka are under way . as an example , diquark breaking mechanisms , absent in the original dpm , have recently been proposed , in the attempt of providing a contribution to the enhanced stopping experimentally observed for nuclear collisions in fixed target experiments . the effects on the prediction of particle / antiparticle asymmetries , which can be compared with measured ones , are under investigation . to test the capabilities of our model , we have simulated hundreds of vertical air shower events , for primary protons and iron nuclei , with energies in the @xmath16 ev range . the average longitudinal eas e/@xmath12 profiles are shown in the upper and lower panels of fig . [ mista14 ] for @xmath17 and @xmath18 ev primary energies , respectively . electrons and positrons of energies @xmath19 1 mev have been considered , together with muons of energies @xmath19 1 gev , up to 1 tev . as can be seen from each figure , at a fixed energy , the @xmath20 depth decreases with increasing primary mass , due to the behaviour of the @xmath15-air inelastic scattering cross - section , which indeed increases with mass number . the same is true also for @xmath21 , whose longitudinal profile is however softer , i.e. has a maximum less peaked than the @xmath22 one . more detail about the @xmath21 longitudinal development can be inferred from fig . [ muvariance ] , where profiles relative to individual showers induced by @xmath15 ( upper panel ) and fe ( lower panel ) primaries at @xmath18 ev are shown . profile fluctuations are more evident for lower mass primaries , also related to the location of the first primary interaction with air , which can even vary within a range of a few km for low mass primaries and low density atmosphere . the muon number reflects the development of the hadronic component of the shower , and is thus a basic test of the hadronic models , since muons come from charged pion and kaon decay . these processes become important at energies below the critical energy , i.e. the energy for which the meson interaction and decay lengths are nearly the same . in particular the critical energy for @xmath23 decay is around 150 gev , while the one for @xmath24 is about ten times larger . from these considerations , it is clear that the @xmath21 number is particularly sensitive to a good description of the hadronic physics processes at low energies . we emphasize that we refer indeed to @xmath21 with energy down to 1 gev . on the other hand , @xmath21 with energies around hundreds or thousands gev in the atmosphere deeply penetrate the rock , and can be selected and detected by underground detectors . multi-@xmath12 and @xmath12-bundle events have also been studied with fluka , which has provided a succesful reproduction of experimental data @xcite . to better identify the longitudinal shape of the shower components , detailed maps of fluences and energy deposition have been built . in fig . [ elec ] and [ muons ] the fluences of @xmath22 and @xmath21 from the top of the atmosphere , down to the earth s surface , are shown for fe induced vertical showers at @xmath25 and @xmath18 ev . from the figures , it is evident that in the top part of the atmosphere the @xmath22 shower profile is more extended than the @xmath21 one , and that some @xmath22 can even propagate from lower altitudes towards higher ones . at the considered energies , some @xmath22 and @xmath21 from the same shower can reach ground within a few km . anyway , the largest fluences are found for both kind of particles in the spatial region located within @xmath14 2 - 8 km above the earth s surface , at distances not larger than 500 - 700 m from the shower axis . in fig . [ emene ] detailed maps of electromagnetic energy deposition for vertical fe showers at @xmath25 and @xmath18 ev are presented . these maps are strictly related to the ones for @xmath26 fluence shown in fig . [ elec ] , already discussed . @xmath22 production thresholds have been set to 1 mev , while @xmath0 threshold has been set to 0.5 mev in our simulations . the ( @xmath27 , @xmath28 ) correlations for @xmath15 and fe showers at sea level and at 4000 m a.s.l . are shown in fig . [ nemu ] and fig . [ highnemu ] , respectively . results for primary energies from @xmath29 to @xmath18 ev are plotted . it appears that for lower energy showers and higher altitudes it is more difficult to distinguish the primary spectrum composition on the basis of ( @xmath27 , @xmath28 ) detection . the features of eas predicted by our simulations are in a qualitative agreement with those obtained by other authors . as a further step in our investigation , we plan to compare in more detail our theoretical results with those obtained by means of other models , e.g. those available in the corsika code . beyond the question of the comparison of hadronic interaction models , this can be helpful to identify also possible differences coming from different e.m . or transport models . this work was partly supported by the university of milano and infn , by the italian murst under contract cofin 2004 , and by department of energy contract de - ac02 - 76sf00515 . 999 d. heck _ et al . _ , report fzka 6019 ( 1998 ) a. ferrari _ et al . _ , cern yellow report 200510 ( 2005 ) , infn / tc_05/11 , 1 - 387 a. fass _ et al . _ , proc . of the chep2003 conference , la jolla , ca , usa , march 24 - 28 , 2003 , ( paper momt005 ) econf c0303241 ( 2003 ) , hep - ph/0306267 s. roesler , r. engel , j. ranft , proc . montecarlo 2000 conference , lisboa , 23 - 26 october 2000 , springer verlag eds . ( 2001 ) , 1033 j. ranft , phys . d * 51 * ( 1995 ) , 64 - 84 g. battistoni _ et al . _ , astropart . phys . * 12 * ( 2000 ) , 315 - 333 g. battistoni _ et al . _ , proc . of the now 2006 workshop , to be published in nucl . suppl . ) . g. battistoni , a. ferrari , m. pelliccioni and r. villari , rad . * 112 * ( 2004 ) , 331 - 343 ; adv . space res . * 36 * ( 2005 ) , 1645 - 1652 s. roesler , w. heinrich and h. schraube , radiation research * * 149 ( 1998 ) 87 ; radiation protection dosimetry * 98 * ( 2002 ) 367 - 388 f. ballarini _ et al . _ , j. of physics : conf . series * 41 * ( 2006 ) , 135 - 142 a. fass _ et al . _ , montecarlo 2000 conference , lisboa , 23 - 26 october 2000 , springer verlag eds . ( 2001 ) , 159 - 164 m. antonelli _ et al . _ , proc on calorimetry in high energy physics ( calor 96 ) , frascati ( italy ) , 814 june 1996 . frascati physics series vol . vi ( 1997 ) , 561 - 570 g. battistoni _ et al . instr . meth . * a394 * ( 1997 ) , 136145 a. ferrari and p.r . sala , proceedings of workshop on nuclear reaction data and nuclear reactors physics , design and safety , a. gandini , g. reffo eds . , trieste , italy , april 1996 , * 2 * ( 1998 ) , 424 - 532 a. capella , u. sukhatme , c.i . tan , j. tran thanh van , phys . rep . * 236 * ( 1994 ) , 225 g. collazuol _ et al . _ , nim a * 449 * ( 2000 ) , 609 . j. ranft , s. ritter , acta phys . b * 11 * ( 1980 ) , 259 ; s. ritter , comput . * 31 * ( 1984 ) , 393 g. battistoni _ et al . _ , 11th int . conf . on nuclear reaction mechanisms , varenna , 12 - 16 june 2006 , e. gadioli ed . ( 2006 ) , 483 - 495 m.g . et al . _ , * b732 * ( 2006 ) , 1 - 45 c. alt _ et al . _ , arxiv : hep - ex/0606028 t. sjstrand , cern report cern / th 7112/93 ( unpublished ) . a. ferrari _ et al . _ , z. phys . * c70 * ( 1996 ) , 413 - 426 a. ferrari_et al . _ , z. phys . * c71 * ( 1996 ) , 75 - 86	fluka is a general purpose monte carlo transport and interaction code used for fundamental physics and for a wide range of applications . these include cosmic ray physics ( muons , neutrinos , eas , underground physics ) , both for basic research and applied studies in space and atmospheric flight dosimetry and radiation damage . a review of the hadronic models available in fluka and relevant for the description of cosmic ray air showers is presented in this paper . recent updates concerning these models are discussed . the fluka capabilities in the simulation of the formation and propagation of em and hadronic showers in the earth s atmosphere are shown .
in recent years , experimental cosmology has provided strong evidence that the universe is currently undergoing a phase of accelerated expansion . the physical phenomena responsible for this expansion has proved elusive , and there are numerous theories that attempt to explain the current epoch of the universe . typically , the observed acceleration is attributed to a new energy component , dark energy , which dominates at late times . numerous models exist of this form , the simplest of which include introducing a cosmological constant @xmath1 , or alternatively a scalar field slowly rolling down a potential . although dark energy is undoubtedly the most popular explanation of the current epoch of the universe , it is not the only way to obtain late time acceleration . in this paper we consider a class of models where gravity is modified at large scales , in such a way that the late - time expansion of the universe could arise naturally from the new gravitational field equations . these modified gravity models have been considered by numerous authors , see for example @xcite . modifying gravity in a manner consistent with experimental data has proved difficult . general relativity ( gr ) is a very robust and well tested theory , and it has been found that even slight modifications often lead to instabilities , such as the propagation of ghosts @xcite , or in otherwise stable matter sources @xcite,@xcite . when introducing additional terms into the gravitational action , we must be careful to respect the success of gr in both the low and high curvature regimes , to ensure that any potentially new model agrees with known observational tests of gravity . in the literature , there are numerous simple conditions that exist which restrict the form that a modified gravity model may take . in this paper , we combine all of these conditions , and in doing so find that the modified gravity function @xmath0 is well constrained , even before considering stringent experimental tests . the philosophy adopted throughout is that before attempting to reconcile any potential modified gravity model with experimental data , that modified gravity model must first satisfy these consistency requirements . we will take a trial @xmath0 model which satisfies all of the existing constraints . we find that our model exhibits a late - time accelerating solution , in the absence of a cosmological constant . we then consider the cosmology of our model , and obtain a modified friedmann equation . we show that to satisfy all constraints , and to obtain an acceptable newtonian limit for small @xmath2 , our model must act like gr with highly suppressed corrections throughout its cosmological history . in addition , we find that although our model has no true cosmological constant , we must still fine tune a particular combination of the free parameters in @xmath0 to obtain an acceptably small curvature in the current accelerating epoch . we conjecture that our results are generic features of models which satisfy the consistency requirements . before we consider a specific @xmath0 model , we first briefly review the field equations in modified gravity theories . throughout this paper , we will study the action @xmath3 where @xmath0 is some function of the ricci scalar @xmath2 , @xmath4 is the mass scale of our model , and @xmath5 the matter action . the field equations obtained by varying the action ( @xmath6 ) with respect to the metric are @xmath7 where @xmath8 is the energy momentum tensor of any matter present . taking the trace of ( @xmath9 ) gives the more useful equation @xmath10 where the function @xmath11 is given by @xmath12 . vacuum solutions to the field equations satisfy @xmath13 when @xmath14 , and hence @xmath15 . in addition to the above field equations in the jordan frame , we will also want to consider the potential of the scalar field in the einstein frame . to transform to the einstein frame , we first write ( @xmath6 ) as @xmath16 + \int \sqrt{-g } d^{4}x{\cal l}_{\rm m } , \ ] ] where @xmath17 is the matter lagrange density . the actions ( @xmath18 ) and ( @xmath6 ) are equivalent if @xmath19 . we then use the conformal transform @xmath20 , which gives @xmath21 where @xmath2 and @xmath22 are the ricci scalars associated with the metrics @xmath23 and @xmath24 respectively . using this in ( @xmath18 ) gives the following einstein frame action @xmath25 + \int \sqrt{-\tilde{g } } d^{4}x { 1 \over ( f')^{2}}{\cal l}_{\rm m } , \ ] ] where the einstein frame potential is given by @xmath26 finally , by making the field redefinition @xmath27 , we obtain @xmath28,\ ] ] with potential @xmath29e^{-2 \sigma\sqrt{2 } / \sqrt{3m^{2}}}.\ ] ] in many instances , it is preferable to study modified gravity in the einstein frame , since the field equations are much simpler . additionally , the potential ( @xmath30 ) is useful since its critical points correspond to the vacuum solutions @xmath31 in the jordan frame . we stress that the two frames are globally equivalent only if @xmath19 . here we briefly review the constraints in the literature imposed on the form of the function @xmath0 . we begin by demanding @xmath32 and @xmath33 for all @xmath2 . the first of these is to ensure stability ; it has been shown @xcite that if @xmath34 for any @xmath2 then the model will possess ghost instabilities in this region . the second condition @xmath19 is imposed since it has been found that models which have @xmath35 will generically possess instabilities , when we consider the metric variational approach of our model @xcite . we note that the possibility of @xmath36 being zero for some @xmath2 has not been completely discounted in the literature . nevertheless , we will impose @xmath33 for all @xmath2 . the next constraint that we impose on @xmath0 is that in the limit @xmath37 , we have @xmath38 . writing @xmath39 , this implies that @xmath40 as @xmath41 . this in turn implies that @xmath42 as @xmath43 , from which we can deduce that @xmath44 as @xmath45 . this condition is imposed to ensure that the modified gravity model acts suitably like @xmath46 in the large @xmath2 regime , where gravity is well constrained by ppn constraints and the cmb . with this bound , we have @xmath42 as @xmath43 . however , in ref . @xcite it was found that for a given @xmath0 theory to be stable to perturbations in the large @xmath2 regime , we must have @xmath47 as @xmath43 . this means that @xmath48 must asymptote to unity from below , since @xmath49 implies that it must be an increasing function . next , we can use the fact that we have imposed @xmath19 for all @xmath2 . this means that there can be no turning points in @xmath48 , implying that if @xmath48 asymptotes to unity from below , then it must remain less than unity for all @xmath2 . however , we have also imposed @xmath50 . from this we conclude that @xmath48 must be an increasing function of @xmath2 , and must satisfy @xmath51 . another set of constraints on @xmath0 are discussed in ref . @xcite , where the cosmology of @xmath0 models was considered . in order for a theory to be a viable alternative to gr , it must exhibit an acceptable cosmological history , that is it must have suitable radiation , matter and late time accelerating eras . it was found that in order for a particular @xmath0 to have a valid matter era , @xmath0 must satisfy @xmath52 and @xmath53 at @xmath54 , where @xmath55 and @xmath56 are given by @xmath57 and @xmath58 . in addition , to obtain a late time accelerating era such as we are currently in , the function @xmath59 must satisfy either ; @xmath60 @xmath61 , @xmath62 and @xmath63 or @xmath64 @xmath65 at @xmath66 . if these are not satisfied for any @xmath2 , then the model will not have an acceptable cosmological history . in this section , we give an example @xmath0 function which satisfies all of the conditions in the previous section . given that @xmath48 must be an increasing function , and lie in the range @xmath67 , an obvious choice is @xmath68,\ ] ] where @xmath69 and @xmath70 are free parameters . this function has the appropriate behaviour @xmath42 as @xmath43 and @xmath71 as @xmath72 . integrating this @xmath48 gives us @xmath73 + a , \ ] ] where @xmath74 is an integration constant . we are looking for @xmath0 models that exhibit late time accelerating solutions in the absence of a cosmological constant . therefore to specify @xmath74 , we impose that @xmath75 . hence we find @xmath76 and @xmath77.\ ] ] it is this @xmath0 that will be studied for the remainder of this paper . we will begin by considering some of the general properties of our model . first , we look for vacuum solutions to the field equations , that is solutions to @xmath15 . the behaviour of @xmath11 is exhibited in fig . we have found that @xmath11 for our @xmath0 generically possesses three zeros , for any @xmath79 and @xmath80 . this means that our @xmath0 model can have three vacuum states , which are minkowski space ( @xmath81 ) , and two de - sitter vacua , one of which is stable . to see these vacuum states more explicitly , we move to the einstein frame . a plot of the potential @xmath82 for our @xmath0 is given in figs . @xmath83 , @xmath84 . we can clearly see that the potential has two minima , one at @xmath85 ( which corresponds to the minkowski vacuum ) and the other at some @xmath86 . this second minima corresponds to a stable de sitter vacuum solution in the jordan frame , and hence our model has a stable accelerating solution , in the absence of a cosmological constant . this is exactly the behaviour we seek . finally , the third vacuum in the jordan frame corresponds to the maxima in the potential @xmath83 , and is unstable . following @xcite , we also briefly consider the functions @xmath55 , @xmath56 and @xmath87 for our model , as given in section @xmath88 . we plot these functions parametrically as functions of @xmath2 in fig . we see that there is a region in which @xmath90 and @xmath91 for @xmath92 , which corresponds to a matter era . as @xmath43 , we have @xmath93 and @xmath94 , suggesting that the condition for a matter epoch is satisfied asymptotically . further , as @xmath2 decreases , there is a region where @xmath95 , at which @xmath96 . this region corresponds to the late - time de - sitter phase . hence our model connects the points @xmath97 and @xmath98 in the @xmath99 plane , as is required @xcite for it to have an acceptable cosmic history . we now consider the newtonian limit of our model . in doing so , we will be able to constrain the free parameters @xmath79 and @xmath70 . the constant @xmath79 will be fixed by the following two conditions ; we require the de - sitter vacuum solution to be @xmath100 , and as we will see shortly , we require @xmath101 . using these conditions , the constant @xmath70 will not be fixed , but we will find that it must be large . to see how @xmath79 and @xmath70 are determined , we consider the newtonian limit of modified gravity theories . the full field equations for the action ( @xmath6 ) are @xmath102 we would like to perform a weak - field expansion of these equations around the de - sitter minima of our model . in doing so , we hope to obtain an acceptable newtonian limit . an analysis has been performed previously in ref . @xcite , and expanding the field equations about a symmetric de - sitter vacuum gives @xmath103 where @xmath104 in order to obtain an acceptable newtonian limit from these perturbative equations , we require that @xmath105 , @xmath106 and @xmath107 , where @xmath108 is a macroscopic distance . if this is the case , then the perturbative equations ( @xmath9 ) will be approximately given by @xmath109 equation ( @xmath110 ) is simply the weak field expansion of general relativity , from which the standard newtonian limit is recovered . the only difference is the coefficient of the first term in ( @xmath110 ) , which is @xmath105 as opposed to unity . note that to obtain the correct limit we must have @xmath111 ; if this were not the case then fourth order derivative terms such as @xmath112 would be present in ( @xmath110 ) , and it has been shown that these terms would lead to strong curvature at all lengthscales @xcite . for our model ( and generically in @xmath0 theories ) , we have @xmath113 for @xmath114 , which is the first condition to obtain an acceptable newtonian limit . however , the second condition , @xmath115 , is not generically satisfied , and we must choose our free parameters to obtain the correct behaviour . from ( @xmath116 ) , we find @xmath36 to be @xmath117 and hence to ensure that @xmath118 is very small , we require @xmath119 . we can use this condition to fix @xmath79 in terms of @xmath120 and @xmath70 . to do so , we note that in the vicinity of the de - sitter minima , @xmath121 , we can expand @xmath0 and @xmath48 for large @xmath122 , @xmath123 and hence from the vacuum field equations @xmath15 , we can use our approximations ( @xmath124 ) to give @xmath125 from which we find @xmath126 . this fixes the free parameter @xmath79 in terms of @xmath120 and the remaining free parameter @xmath70 . next , we can constrain @xmath70 from the condition @xmath127 . using this , and expanding @xmath36 in powers of @xmath128 , we find that @xmath70 must satisfy @xmath129 which is satisfied for large @xmath70 , such that @xmath130 . to review our conditions ; we are working with a model with two free parameters , @xmath79 and @xmath70 , and we have imposed two constraints on this model , which are that @xmath131 and @xmath132 . these constraints force us to choose @xmath79 and @xmath70 such that @xmath133 and @xmath134 . together , they imply that @xmath135 , and hence @xmath136 . we note that we have not been forced to specify the constant @xmath70 , other than imposing the condition @xmath135 . this parameter controls how closely our model mimics gr ; the larger we set @xmath70 the more suppressed our corrections @xmath137 become . before moving onto the cosmology of our model , we remark that the above weak - field expansion was performed around the de - sitter vacuum @xmath138 . it has been noted @xcite that for local tests of gravity , the expansion should instead be around the local energy density of the solar system , which is much larger than @xmath120 . in a subsequent paper we will perform a detailed analysis of the local tests of gravity for our model , but for now we simply expand around @xmath139 . performing a weak - field expansion around a solar system background will not significantly modify our conclusion ; we are simply stating that in order to obtain an acceptable weak - field limit for small @xmath2 , we must choose the constants @xmath79 and @xmath70 such that @xmath140 is satisfied . having considered the newtonian limit of our models , we now evaluate the friedmann equation , and attempt to reconstruct the cosmic history of our models . we will find that this model mimics general relativity very closely throughout its cosmic history . to show this , we first write down the equivalent friedmann equation for a general @xmath0 model , obtained from the @xmath141 component of the field equations ( @xmath9 ) , @xmath142 = { \rho \over m^{2 } } .\ ] ] with our highly non - trivial @xmath0 , this equation corresponds to a complex , non linear , third - order differential equation in the scale factor @xmath79 . however , we can use our approximations ( @xmath124 ) to write ( @xmath143 ) in a much simplified form . we first note that the approximations ( @xmath124 ) are valid for the entire cosmic history of this model , because they are applicable for large @xmath144 . since @xmath145 throughout the cosmic history , and since @xmath146 is at its smallest value at @xmath138 , then given that we have constrained @xmath146 to be large at @xmath138 then it must be large for all @xmath145 . using the approximations ( @xmath124 ) in ( @xmath143 ) , we find the following modified friedmann equation , @xmath147 where we have used the condition that @xmath70 must be large , and hence @xmath148 . the first two terms are what we would expect from standard general relativity with a fine tuned cosmological constant @xmath149 . the remaining terms on the left hand side are corrections to the friedmann equation , but they are highly suppressed by a factor of @xmath150 for all @xmath151 . it follows that our model will not significantly differ from standard gr . the exponentially suppressed corrections to general relativity that we have found are small in the current epoch , and decrease in significance as we track our model back through its cosmic history . additionally , we note that we still have a fine tuning problem , in that we have two free parameters , and we have chosen them specifically such that the combination @xmath152 satisfies @xmath153 . we stress however that the combination @xmath154 in ( @xmath155 ) is not a true cosmological constant ; rather our model mimics standard general relativity with a small vacuum energy . the fine - tuning found in our model is a generic feature of modified gravity theories . since we demand that @xmath156 for small @xmath2 to obtain an acceptable weak field limit , it follows that in the vicinity of the vacuum @xmath157 , we must have @xmath158 , and hence @xmath159 , where @xmath160 and @xmath161 are constants . this means that for vacuum @xmath121 , we can expand @xmath0 in the form @xmath162 , where @xmath160 and @xmath161 will be some combination of constants in the model . this is exactly what we have done for our @xmath0 model above ; we have expanded @xmath0 as @xmath163 for @xmath121 . it follows that to obtain an acceptable vacuum solution @xmath164 , we must fine - tune the parameters in the model . the fact that our model must necessarily mimic gr with suppressed corrections is a consequence of the constraints that we have imposed . specifically , we have used the fact that to obtain an acceptable weak field expansion , we require @xmath156 for small @xmath2 . in addition , we have also assumed that in the large @xmath2 regime @xmath165 , which suggests that @xmath156 as @xmath41 . since we are considering models which have @xmath19 for all @xmath2 , then it follows that @xmath36 must be small throughout the cosmic history . we expect that any @xmath0 function that satisfies the consistency conditions that we have considered will possess a friedmann equation of the form ( @xmath155 ) , that is standard gr plus @xmath1 with small correction terms . these correction terms will not necessarily be exponentially suppressed as they are for the @xmath0 considered in this paper , but they must be suppressed throughout the expansion history of the universe regardless of the particular form of @xmath0 . however , although these models will mimic gr , this does not imply that they are observationally identical to gr if we consider all aspects of gravity . for example , it might be possible to observe small corrections to gr by considering solar system tests of gravity or the binary pulsar . in a subsequent paper we hope to find distinct observational signatures of our model by analyzing local gravity constraints . finally , we observe that the conclusions that we have reached do not apply if we were to relax any of the constraints on @xmath0 , for example by allowing @xmath166 for some @xmath2 in the past . hence it might be possible to construct an acceptable model that does not obey all of the constraints at all times , and hence will not mimic gr throughout the entire cosmic history . in addition , it is not strictly necessary for an acceptable modified gravity model to obey the consistency conditions for all @xmath2 , since they can only be probed for @xmath167 . we could have an acceptable model which has @xmath168 for some @xmath169 in the future . however , there is a possibility of such models evolving to a final state containing instabilities . we have taken the consistency constraints that exist in the literature for modified gravity models , and used them to construct a particular modified gravity function @xmath0 . a study of the model has shown that in order to satisfy all constraints , it must act like gr with highly suppressed corrections throughout cosmic history . in addition , we find that even though we do not have a true cosmological constant in our setup , we must still fine - tune the parameters in our @xmath0 function to obtain an acceptable late time accelerating epoch . although we have come to this conclusion for our specific model , we expect our results to hold true for a large class of @xmath0 models which satisfy all of the constraints that we have considered . while we were completing this work we became aware of refs . @xcite , which come to similar conclusions . in ref . @xcite , it was found that in order to satisfy local gravity constraints , the function @xmath55 as defined in section @xmath88 must satisfy @xmath170 for @xmath171 . this places a severe constraint upon @xmath36 , which must be very small in the current epoch of the universe . the model considered in this paper can satisfy the constraint @xmath170 by making the free parameter @xmath70 sufficiently large . however , such a stringent constraint is likely to yield a model that is virtually indistinguishable to gr . was considered , where @xmath59 , @xmath173 , @xmath174 and @xmath175 are free parameters . the model considered in this paper shares many similarities with ( @xmath176 ) ; the condition @xmath177 was imposed in ref . @xcite both at @xmath178 and as @xmath43 . this implies that the function @xmath179 can be expanded as for small @xmath181 , which is valid throughout the cosmic history of the model . we have arrived at a very similar expansion for the @xmath0 function considered in this paper , ( @xmath124 ) . the difference between the two models is that ( @xmath182 ) contains power law corrections to gr , whereas ( @xmath124 ) contains exponentially suppressed corrections .	modified gravity models are subject to a number of consistency requirements which restrict the form that the function @xmath0 can take . we study a particular class of @xmath0 functions which satisfy various constraints that have been found in the literature . these models have a late time accelerating epoch , and an acceptable matter era . we calculate the friedmann equation for our models , and show that in order to satisfy the constraints we impose , they must mimic general relativity plus @xmath1 throughout the cosmic history , with exponentially suppressed corrections . we also find that the free parameters in our model must be fine tuned to obtain an acceptable late time accelerating phase . we discuss the generality of this conclusion .
the sense of hearing requires exquisite mechanical detection , with barely audible tones evoking displacements of the basilar membrane on the order of angstroms @xcite . the auditory system is also highly tuned , with frequency selectivity in various species reaching 0.1% , and the frequency range reaching as high as 100 khz . over the last 60 years , there have been significant advances in our understanding of the inner ear . however , the detailed mechanisms of the auditory system are still not understood , and thus , deficits are being mostly aided by technological solutions , such as cochlear implants . nonlinear effects have been shown to be important for the extreme sensitivity and robustness of the inner ear @xcite . compressive nonlinearity plays a role both in protecting the cells from damage , and for ensuring that the lowest levels of incoming sound receive the highest degree of amplification @xcite . nonlinear response has been demonstrated both at the organism level @xcite and in the motility of individual hair cells @xcite . here , we focus on the latter , since hair cells constitute the main functional elements in the detection process . on the apical surface of the hair cell , 20 - 300 stereocilia comprise the hair bundle ; tips of neighboring stereocilia are connected by tip links @xcite . during stimulus , deflections due to incoming sound induce shearing of the stereocilia comprising the hair bundle , increasing the tension on the tip links between them . the links are coupled to mechanically sensitive ion channels , which open and close in response to the stimulus forces @xcite . the resulting influx of ions depolarizes the cell , and thus leads to the release of neurotransmitters . when the bundles are deflected by sound waves , they move in a highly viscous medium . an active process has therefore been proposed to explain the high acuity of hearing @xcite . several models were developed to explain how the hair cell generates the forces needed to pump energy into the oscillation . these include amplification by active hair bundle motility @xcite and electromotility , a process of elongation and contraction of the hair cell soma in response to electrical stimulation @xcite . empirical evidence provided by otoacoustic emissions @xcite indicates an underlying active mechanism in the auditory response . synchronization of hair bundle oscillations by periodic perturbations , exhibited over a wide range of frequencies @xcite , suggests that the auditory system is analogous to general forced oscillatory media , similar to autocatalysis in chemical media or enzymatic dynamics in physiology @xcite . to explain the role of active amplification in hearing , theoretical models proposed that hair cell response follows temporal dynamics that arise through a _ hopf _ bifurcation @xcite , a generic mechanism that describes the birth of oscillatory behavior once the critical value of a control parameter is exceeded @xcite . the model equations predict high gain and sharp frequency selectivity at low - amplitude stimuli , and a reduction of both with increasing amplitudes . the amplification gain was shown to diverge as the control parameter approaches a critical value and to diminish away from the critical point @xcite . consequently , proximity to a hopf bifurcation is recognized to provide important advantages in explaining the phenomenology of hearing @xcite . experimental studies have shown that the dynamic response of a hair cell intertwines many degrees of freedom and complex processes , including biochemical feedback on the control parameter @xcite . yet , theoretical models have typically included only additive forcing terms @xcite . we develop a general theoretical framework that allows a systematic study of the impact of parametric versus additive forcing on the resonant response in the cochlea . the two types of forcing reflect the coupling between the driving force and the original ( unforced ) system , with the parametric term reflecting a situation in which the periodic forcing directly impacts one or more parameters . for example , the light - sensitive oscillatory belousov zhabotinsky chemical reaction under periodic illumination is a parametrically forced system , since the light affects a chemical reaction @xcite . the oscillatory nature of the belousov zhabotinsky chemical reaction is analogous to the spontaneous hair bundle oscillations , while the role of illumination @xcite is analogous to feedback by calcium ions . frequency locking is a generic feature of periodically driven oscillatory systems that exhibit resonant behavior , examples of which include faraday waves , nonlinear optical solitons , josephson junctions , and chemical reactions . frequency locked response has also been shown to be a crucial feature of auditory detection @xcite . we examine the properties of the 1:1 resonance domain ( arnold tongue ) in the cochlear response . we consider hair cells to be poised in the vicinity of the hopf bifurcation @xcite and derive a universal normal form equation that includes both additive and parametric driving forces . specifically , we examine the distinctions in the transition from unlocked to locked oscillations , under different types of forcing . this model of the cochlear response allows for the coexistence of multi - modal frequency locking ( i.e. , beyond 1:1 resonance ) and elucidates the presence of super- vs. sub - critical forms of the frequency locking transition @xcite . thus , this study provides a framework for incorporating both biochemical and mechanical feedback in the description of the auditory system . periodically forced oscillatory systems can be mathematically represented as follows : @xmath1 where @xmath2 is a set of observables ( with @xmath3 being an integer ) , @xmath4 and @xmath5 denote functions that describe interactions between observables , and @xmath6 is the frequency of the driving force . in what follows , we will consider , without a loss of generality , the two variable activator inhibitor fitzhugh - nagumo ( fhn ) model @xcite , in which @xmath7 . forcing acts on the activator variable @xmath8 ; up to linear order , it is given by @xmath9 , and @xmath10 . we note that higher order forcing terms do not contribute to the discussed results , which focus on 1:1 resonance . magnitudes of the additive and parametric forcing terms are represented by @xmath11 and @xmath12 , respectively . additional details on the fhn model are given in the appendix . near the hopf bifurcation and in the absence of periodic forcing , model equations of type can be reduced to a universal nonlinear equation , which is often referred to as the stuart - landau ( a.k.a . complex ginzburg - landau ) equation @xcite . under externally applied periodic forcing ( at frequency @xmath6 ) , the phase invariance along the limit cycle is destroyed , and the system exhibits either unlocked oscillations or entrained oscillations with discrete phase shifts . increasing the amplitude of the drive increases the range of detuning under which a resonant solution can arise . this detuning from an unforced characteristic hopf frequency @xmath13 is given by @xmath14 and thus corresponds to @xmath15 frequency locking ( where @xmath16 is an integer ) , for which the original system responds at translations @xmath17 @xcite . under small detuning values @xmath18 , the stuart - landau equation is modified as @xcite : @xmath19 where @xmath20 is a complex amplitude that describes weak temporal modulations of a primary limit cycle that is generated at the hopf onset , @xmath21 measures the distance from the hopf bifurcation , @xmath22 is the nonlinear frequency correction , @xmath23 is a ( real ) forcing magnitude , and @xmath24 is the complex conjugate of @xmath20 . notably , the @xmath15 resonant solutions are invariant under @xmath25 @xcite ; hence , the frequency locking condition implies that @xmath26 is constant . this resonance condition is fulfilled over a finite range of driving frequencies and amplitudes , which defines the domain of an arnold tongue . we focus this study on the 1:1 resonant response @xcite . the amplitude equation ( [ eq : fcgl ] ) with @xmath27 has been employed in several contexts , such as studies of fluctuations and the response to pitches @xcite . here , we examine the physical nature of the forcing and its implications for the resonant response . in what follows , we show that for 1:1 resonance , one can obtain a generalized amplitude equation @xmath28 and explore its frequency locking properties ; for details , we refer the reader to the appendix . to determine the regions of frequency locked solutions , we rewrite and separate the contributions of the additive and parametric components by the transformation @xmath29 next , we introduce the following notation to distinguish the additive and the parametric forcing terms : @xmath30 for which : @xmath31 using the polar representation @xmath32 and looking for stationary solutions , we impose the conditions : [ eq : amp_rho_phi ] @xmath33 solutions to ( [ eq : amp_rho_phi ] ) thus satisfy an equation for the amplitude @xcite : @xmath34 and for the phase [ eq : amp_phi ] @xmath35 for the additive case ( @xmath36 ) , the phase displays @xmath37 symmetry shifts , while for the parametric case ( @xmath38 ) , one obtains phase shifts of @xmath0 . the latter implies bistability of frequency locked solutions , viz . @xmath39 and @xmath40 , where @xmath41 and @xmath42 are solutions to ( [ eq : amp_rho_phi ] ) . this result is analogous to a situation where a spatially periodic system is driven by a space - dependent modulation @xcite . thus , the @xmath43 resonant solutions can exhibit different phase symmetries , with one of them displaying properties similar to the @xmath44 resonant solutions @xcite . the linear stability of these solutions @xmath45 is determined by the sign of the real part of the eigenvalues @xmath46},\ ] ] where the solutions are stable if @xmath47 and unstable otherwise . combining the results on existence and stability of frequency locked solutions , we obtain the resonance regimes , for parameter space that is spanned by the forcing amplitude ( @xmath48 or @xmath49 ) and detuning ( @xmath50 ) . in particular , we distinguish between two cases : the oscillatory regime ( @xmath51 ) and the quiescent regime ( @xmath52 ) , as shown in fig . [ fig : above_hopf ] and fig . [ fig : below_hopf ] , respectively . ( a ) ) , in a parameter space of detuning ( @xmath50 ) and forcing magnitudes for ( a ) parametric forcing , @xmath53 and ( b ) additive forcing , @xmath54 . the bottom panel describes the resonant region ( shaded area ) , while the top panels describe the amplitudes of resonant solutions and unlocked oscillations ( light dashed lines ) at two distinct @xmath55 values ( @xmath56 , @xmath57 ) as a function of @xmath50 ; solid lines in top / mid panels mark stable solutions , and the dashed line in the bottom panel of ( a ) marks the locus of points at which the nontrivial solutions bifurcate . equation [ eq : modfcgl ] was solved with parameters : @xmath58 and ( a ) @xmath59 , ( b ) @xmath60 . ] ( b ) ) , in a parameter space of detuning ( @xmath50 ) and forcing magnitudes for ( a ) parametric forcing , @xmath53 and ( b ) additive forcing , @xmath54 . the bottom panel describes the resonant region ( shaded area ) , while the top panels describe the amplitudes of resonant solutions and unlocked oscillations ( light dashed lines ) at two distinct @xmath55 values ( @xmath56 , @xmath57 ) as a function of @xmath50 ; solid lines in top / mid panels mark stable solutions , and the dashed line in the bottom panel of ( a ) marks the locus of points at which the nontrivial solutions bifurcate . equation [ eq : modfcgl ] was solved with parameters : @xmath58 and ( a ) @xmath59 , ( b ) @xmath60 . ] ( a ) ) , in a parameter space of detuning ( @xmath50 ) and forcing magnitudes for parametric forcing , @xmath53 . the bottom panel describes the resonant region ( shaded area ) , while the top panel describes a typical behavior at a specific @xmath55 ( @xmath61 ) value as a function of @xmath50 ; solid lines in the top panel mark stable solutions , and the dashed line in the bottom panel marks the locus of sub - critical bifurcation onsets for the nontrivial solution . ( b ) a typical amplitude dependence at a specific @xmath55 value ( @xmath62 ) as a function of @xmath50 , for the additive case ( dark line ) , with a superimposed parametric case ( light line ) that is taken from ( a ) at slice ( i ) . equation [ eq : modfcgl ] was solved with parameters : ( a ) @xmath63 , @xmath59 and ( b ) @xmath64 , @xmath60 . ] ( b ) ) , in a parameter space of detuning ( @xmath50 ) and forcing magnitudes for parametric forcing , @xmath53 . the bottom panel describes the resonant region ( shaded area ) , while the top panel describes a typical behavior at a specific @xmath55 ( @xmath61 ) value as a function of @xmath50 ; solid lines in the top panel mark stable solutions , and the dashed line in the bottom panel marks the locus of sub - critical bifurcation onsets for the nontrivial solution . ( b ) a typical amplitude dependence at a specific @xmath55 value ( @xmath62 ) as a function of @xmath50 , for the additive case ( dark line ) , with a superimposed parametric case ( light line ) that is taken from ( a ) at slice ( i ) . equation [ eq : modfcgl ] was solved with parameters : ( a ) @xmath63 , @xmath59 and ( b ) @xmath64 , @xmath60 . ] above the onset of the hopf bifurcation , both the parametric and the additive forcing terms lead to a similar arnold tongue response , as shown by the shaded domains in fig.[fig : above_hopf ] . at weak forcing magnitudes ( @xmath65 ) , the frequency locked solutions form isolas and coexist with an additional unstable solution ; in the parametric case , this unstable state is a trivial one . the shaded region is identified with solutions that are linearly stable with respect to temporal perturbations . outside of the resonance domain , unlocked oscillations prevail , with the maximal amplitude denoted by the light dashed line , as shown in the horizontal slices ( i ) at fixed @xmath55 . at larger values of @xmath12 , two new solutions bifurcate from the trivial state ( see the dark dashed lines in the resonance domain of fig . [ fig : above_hopf](a ) ) , where the right line denotes solutions of a super - critical nature , while the left is sub - critical ; both are characteristic of the 2:1 resonance @xcite . the left non - trivial solution continues towards negative detuning values , folds back toward positive values , and connects with the right super - critical branch . stability of this top branch defines the resonant solutions . in the additive forcing case ( @xmath11 ) , the isola merges with the bottom solution through a cusp bifurcation @xcite , which results in a single amplitude throughout the whole detuning region ( not shown here ) . both behaviors are described by slices ( ii ) at fixed @xmath55 . further details on the coexistence and stability of such solutions are of secondary significance and will be discussed elsewhere . properties of the frequency locked solutions in the quiescent regime are fundamentally different from those of the innately oscillatory regime . in the parametric case , the resonant solutions also take the form of an arnold tongue . however , outside of the resonance region , the trivial state is stable , and thus the system can be either quiescent ( outside of the resonance regime ) or frequency locked ( within the resonance regime ) . nevertheless , since the resonance boundaries preserve the super- and sub - critical properties , part of the resonance boundary exhibits _ hysteresis _ , as shown in fig . [ fig : below_hopf](a ) . these properties imply a smooth transition to frequency locked oscillations if one approaches the arnold tongue from large positive detuning , and an abrupt transition if the resonance is approached from negative detuning values . the former case is reversible upon detuning , while the latter is associated with two distinct transition onsets . the frequency locked solutions in the additive case persist throughout the whole parameter range , as shown in fig . [ fig : below_hopf](b ) by a typical slice along @xmath55 . nevertheless , even in the absence of a distinct transition to resonance , we can identify amplification of the response amplitude . notably , while the resonant behavior is smooth for the fhn model , it is possible to observe a hysteresis here as well , via a cusp bifurcation @xcite . however , the hysteresis designates a transition from one oscillatory state to another , rather than a transition from the quiescent state , as in the parametric case . hair cells were shown to be the sources of amplification in the inner ear @xcite , operating either via hair bundle motility @xcite or somatic electromotility @xcite . according to theoretical models , the amplification gain and frequency selectivity of the auditory response are dependent on the value of an internal control parameter . as empirical evidences support the existence of a feedback mechanism , which modulates this control parameter in response to external forcing @xcite , it is important to incorporate its dynamics into the theoretical models . a universal framework has therefore been formulated to describe the 1:1 frequency locking of a system poised near the hopf bifurcation and exposed to both additive and parametric forcing . while additive forcing ( @xmath11 term in ( [ eq : modfcgl ] ) ) has been employed in previous studies , we incorporate here a periodic forcing term that couples directly to biochemical processes ( @xmath66 term in ( [ eq : modfcgl ] ) ) modulating the internal control parameter . a related study of synchronization dynamics of coupled oscillators also supports the inclusion of biochemical feedback @xcite . we discuss below the empirical data indicating the influence of both types of forcing on the frequency locked response . for the 1:1 resonance , we explored the scaling of the response amplitude with the additive and parametric forcing magnitudes . as a result , we obtain responses that are characterized by distinct power laws . a number of power laws have been experimentally observed in the response of the auditory system and are discussed in @xcite . the experimental results show three distinct regimes in the response : ( i ) at low forcing magnitudes , the response scales linearly , ( ii ) as the magnitude is increased , there is a crossover to a nonlinear regime with @xmath67 exponent , and ( iii ) an additional transition is observed at high forcing magnitudes . the additive forcing can explain the first two cases , which result from the competition between nonlinear and linear terms . however , the crossover to linearity in the third regime @xcite is not captured by including only the additive forcing term . the response to parametric forcing exhibits linear scaling @xmath68 , and hence supports the emergence of the third regime . most of the model equations have employed additive forcing , where the 1:1 resonant solutions obey @xmath69 symmetry @xcite . the generalized equations developed here show that the inclusion of a parametric forcing term introduces bistability of solutions differing by phase shifts of @xmath0 . phase shifts of @xmath0 have apparently been observed in experiments @xcite , but not accounted for in a theoretical model . construction of resonance domains ( i.e. , arnold tongues ) provides insight into the transition from unlocked to frequency locked dynamics . indeed , experimental measurements have been performed and revealed complex dynamics in the transition from spontaneous oscillation into the resonance regime @xcite . experimental results have raised conjectures on the possible presence of both super- and sub - critical forms of the hopf instability @xcite . our results demonstrate that a hysteresis can be described solely by bifurcations of the phase - locked states , which leave the super - criticality of the hopf bifurcation intact . resolving the contributions from the additive versus parametric forcing is difficult in the regime above the hopf onset ( spontaneously oscillatory regime ) , as the magnitude of unlocked oscillations is equal to that of the locked states . however , the differences become clear in the regime at or slightly below the hopf onset , where the hysteresis is conjectured to occur in the transition from a quiescent to a frequency locked state . the results presented here are not limited to a specific model and should arise as general features of 1:1 frequency locking @xcite in other systems , such as elastically driven cardiomyocytes @xcite . we believe that this framework provides a more realistic description of the biological system , as it can incorporate both biochemical and mechanical feedback in descriptions of the auditory response . further , the theoretical framework that incorporates parametric behavior could be generalized to include the coexistence of multiple @xmath15 resonances , which have not been included in previous spatially extended models . hence , the methodology developed here could provide a framework for future spatiotemporal models of the cochlear response . finally , this study of frequency locking dynamics can be generalized to other systems , such as faraday waves , shaken granular media , forced oscillatory chemical reactions , and elastically forced cardiomyocytes . we thank robijn bruinsma , oreste piro , and ehud meron for helpful discussions on the subject . this work was partially supported by the adelis foundation ( a.y . ) , and by nih grant r01dc011380 , nsf grant ios-1257817 ( db ) . the fitzhugh - nagumo model is a general bonhoeffer - van der pol type equation , which has been employed as a prototypical model for many biological and chemical systems @xcite . the response of the system to periodic forcing is described by : where @xmath8 is an activator , @xmath71 is an inhibitor , and @xmath72 and @xmath73 are parameters . we note that the bonhoeffer - van der pol type equations have already been employed in modeling the dynamics of the auditory system @xcite . the trivial solution to the unforced equation ( [ eq : fhn ] ) crosses the hopf instability at @xmath74 and a critical frequency @xmath75 . near the instability onset , @xmath76 , and under 1:1 periodic forcing , eqs . [ eq : fhn ] obey the approximation : @xmath77 where @xmath78 employing standard multiple time - scale expansion , and letting @xmath79 and @xmath80 @xcite , we obtain , up to order @xmath81 , the generalized amplitude equation for 1:1 forcing that incorporates both additive and parametric components : @xmath82 where , @xmath83 68 ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1 '' '' @noop [ 0]secondoftwo sanitize@url [ 0 ] + 12$12 & 12#1212_12%12 @startlink[1 ] @endlink[0 ] @bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )	the auditory system displays remarkable sensitivity and frequency discrimination , attributes shown to rely on an amplification process that involves a mechanical as well as a biochemical response . models that display proximity to an oscillatory onset ( a.k.a . hopf bifurcation ) exhibit a resonant response to distinct frequencies of incoming sound , and can explain many features of the amplification phenomenology . to understand the dynamics of this resonance , frequency locking is examined in a system near the hopf bifurcation and subject to two types of driving forces : additive and parametric . derivation of a universal amplitude equation that contains both forcing terms enables a study of their relative impact on the hair cell response . in the parametric case , although the resonant solutions are 1:1 frequency locked , they show the coexistence of solutions obeying a phase shift of @xmath0 , a feature typical of the 2:1 resonance . different characteristics are predicted for the transition from unlocked to locked solutions , leading to smooth or abrupt dynamics in response to different types of forcing . the theoretical framework provides a more realistic model of the auditory system , which incorporates a direct modulation of the internal control parameter by an applied drive . the results presented here can be generalized to many other media , including faraday waves , chemical reactions , and elastically driven cardiomyocytes , which are known to exhibit resonant behavior .
a common symptom of correlated - electron systems is magnetism , and neutron scattering is the premiere technique for measuring magnetic correlations in solids . with a spin angular momentum of @xmath3 , the neutron interacts directly with the magnetization density of the solid . elastic scattering can directly reveal static magnetic order ; for example , neutron diffraction provided the first experimental evidence for nel antiferromagnetism @xcite . through inelastic scattering one can probe dynamic spin - spin correlations ; in an ordered antiferromagnet , one can measure the precession of the spins about their average orientations , which show up as dispersing spin waves . neutrons do not couple to the charge of the electrons , but instead scatter from atomic nuclei via the strong force . despite the name , the small size of the nucleus compared to the electronic charge cloud of the atom results in a rather weak scattering cross section . the magnetic and nuclear scattering cross sections are comparable , so that neutron scattering is very sensitive to magnetism , in a relative sense . a challenge with neutron scattering is that the combination of weak scattering cross section and limited source strength means that one needs a relatively large sample size compared with many other techniques . the value of the information that can be obtained by neutron scattering generally makes worthwhile the effort to grow large samples ; nevertheless , in practice it is useful to take advantage of complementary information obtained from techniques such as muon spin rotation spectroscopy and nuclear magnetic resonance . the latter techniques yield less information but often provide greater precision . there have also been continuing developments in resonant x - ray scattering ; nevertheless , neutron scattering will remain an essential technique to investigate strongly correlated systems for the foreseeable future . as we have space only for a concise introduction to the field , we note that there plenty of more extended references available . a good summary of the theory of neutron scattering is given by squires @xcite , while a more detailed description is provided by lovesey @xcite . we have contributed to a technique - oriented book @xcite and to book chapters on magnetic neutron scattering @xcite , and new books on the subject continue to appear . to illustrate some of the concepts and capabilities , we will use examples involving copper - oxide compounds , especially from the family la@xmath0ba@xmath1cuo@xmath2 , which includes phenomena from antiferromagnetic order to high - temperature superconductivity . more details on neutron scattering studies of cuprates are given in recent reviews @xcite . .basic properties of a neutron . the gyromagnetic ratio , @xmath4 , and the g - factor , @xmath5 , are defined by @xmath6 , where @xmath7 is the neutron s angular momentum , @xmath8 is the neutron s spin ( @xmath9 ) , @xmath10 j / t @xmath11 erg / gs is the nuclear magneton @xcite . [ cols="^,^,^,^,^,^ " , ] [ tranq : tab:2 - 2 ] neutrons are especially abundant in nuclei of high atomic number , where they can significantly exceed the number of protons . to create a neutron beam , the first challenge is to extract neutrons from the nuclei . the first practical source was the nuclear reactor , in which neutron bombardment of @xmath12u nuclei induces fission , a process that releases several neutrons per incident neutron , thus allowing for a self - sustaining chain reaction . the neutrons that are released have a very large energy , whereas the fission cross section is enhanced by slower neutrons . the slowing of neutrons can be achieved quite effectively by scattering from hydrogen , especially in the form of h@xmath13o , which can also act to cool the reactor core . in a research reactor , where one would like to extract some of the neutrons , the reactor moderator can be made more transparent to neutrons by replacing h@xmath13o with d@xmath13o ( heavy water , with d representing deuterium ) . cylindrical thimbles poking into the water moderator provide an escape path for neutrons , which form the beams that supply neutron spectrometers . another approach is to knock the neutrons out of heavy nuclei with high - energy protons from an accelerator . again , the neutrons that can escape the nuclei have very high energies that must be reduced by multiple scattering in a moderator . in contrast to a reactor , which produces neutron beams that are continuous in time , the proton beam provided by an accelerator can be pulsed , so that a spallation source typically has pulsed beams of neutrons . targets can be made of a heavy metal such as tungsten , but newer sources with higher power tend to use liquid mercury in order to allow adequate heat removal . a list of the major operating spallation sources in the world is given in the upper portion of table [ tranq : tb : source ] . information on the available instrumentation and capabilities can be obtained from the listed web sites . with a pulsed neutron source , each burst of neutrons is produced in a narrow time window , so that one can distinguish between neutrons of different velocities by their travel time , or `` time of flight '' . using a rotating shutter , one can select incident neutrons of a desired energy ; the energy of scattered neutrons can then be determined by their time of arrival at a detector . @llll@ facility & laboratory & location & web address + 5 mm _ spallation sources _ + spallation neutron source & oak ridge national lab & oak ridge , tn , usa & http://neutrons.ornl.gov/ + lujan neutron scattering center & los alamos national lab & los alamos , nm , usa & http://lansce.lanl.gov/lujan/ + isis & rutherford appleton lab & didcot , uk & http://www.isis.stfc.ac.uk/ + j - parc & japan atomic energy agency & tokai , japan & http://j-parc.jp/matlife/en/ + sinq & paul scherrer institut & villigen , switzerland & http://www.psi.ch/sinq/ + + 5 mm _ reactor facilities _ + high flux isotope reactor & oak ridge national lab & oak ridge , tn , usa & http://neutrons.ornl.gov/ + nist center for neutron research & nist & gaithersburg , md , usa & http://www.ncnr.nist.gov/ + institut laue langevin & ill & grenoble , france & http://www.ill.eu/ + frm - ii & technische universitt & munich , germany & http://www.frm2.tum.de/en/ + & mnchen & & + laboratoire lon brillouin & cea saclay & saclay , france & http://www-llb.cea.fr/en/ + jrr-3 & japan atomic energy & tokai , japan & http://qubs.jaea.go.jp/en_index.html + & agency & & + opal & ansto & lucas heights , nsw , & http://www.ansto.gov.au/ + & & australia & + hanaro & kaeri & daejeon , south korea & http://hanaro4u.kaeri.re.kr/ + [ tranq : tb : source ] the spallation source sinq at the paul scherrer institut provides a continuous , rather than pulsed , beam , so its instrumentation has more in common with reactor facilities , which are listed in the lower portion of table [ tranq : tb : source ] . with a continuous source , it is common to select the desired energy of incident neutrons by bragg diffraction from a crystal ( or array of crystals ) . in a triple - axis spectrometer @xcite , one also uses bragg diffraction to analyze the energy and momentum of neutrons scattered by a sample . again , many of the facility web sites provide a wealth of information on spectrometers and capabilities . many of the fundamental advantages of neutron scattering techniques arise from the fact that the neutron s interactions with matter are usually weak and are extremely well understood . hence , neutrons afford direct experimental insight into dynamical properties of the material system of interest , unperturbed by the probe and essentially undistorted by the details of its interaction with matter . these properties contrast favorably with x - ray or charged - particle ( electron , muon ) techniques , where the probe could significantly perturb the system , and the interaction matrix elements between the system and the probe are often very complicated and profoundly impact the physics measured in the experiment . the scattering of neutrons by an atomic system is governed by two fundamental interactions . the residual strong interaction ( nuclear force ) gives rise to scattering by the atomic nuclei ( nuclear scattering ) . the electromagnetic interaction of the neutron s magnetic moment with the sample s internal magnetic fields , mainly originating from the unpaired electrons in the atomic shells , gives rise to magnetic scattering @xcite . magnetic interaction of a neutron with a single atom is of relativistic origin and is very weak , so that magnetic neutron scattering can be treated using the born approximation . the interaction potential consists of the dipole - dipole interaction with the magnetic moment associated with the electronic spin , @xmath14 ( @xmath15 is the land @xmath16-factor ) , @xmath17and the interaction with the electric current associated with the electron s orbital motion @xmath18here @xmath19 is the electron s orbital angular momentum , and @xmath20 its coordinate in the neutron s rest frame . while the neutron s interaction with the atomic nucleus is strong the nuclear force is responsible for holding together protons and neutrons in the nucleus it has extremely short range , @xmath21 cm , comparable to the size of the nuclei , and is much smaller than the typical neutron s wavelength . hence , to describe the neutron s interaction with the system of atomic nuclei in which the typical distances are about 1 @xmath22 cm , a highly accurate approximation is obtained by using a delta - function for the nuclear scattering length operator in the coordinate representation , @xmath23here @xmath24 is the position of the nucleus and @xmath25 is the nuclear scattering length , which is usually treated as a phenomenological parameter @xcite that has been determined experimentally and tabulated @xcite . in the born approximation , the scattering length would correspond to the neutron - nucleus interaction described by the fermi s _ pseudo - potential _ @xcite , @xmath26 in general , the bound scattering length ( that is , for a nucleus fixed in space ) is a complex quantity @xcite , @xmath27 , defining the total scattering cross - section , @xmath28 , and the absorption cross - section far from the nuclear resonance capture , @xmath29 , through @xmath30for the majority of natural elements @xmath31 is close in magnitude to the characteristic magnetic scattering length , @xmath32 fm ( 1 fm @xmath33 cm and @xmath34 is the classical electron radius ) . in a scattering experiment , the sample is placed in the neutron beam having a well - defined wave vector @xmath35 and known incident flux density @xmath36 , and the detector measures the partial current , @xmath37 , scattered into a small ( ideally infinitesimal ) volume of the phase space , @xmath38 , near the wave vector @xmath39 , as indicated in fig . [ tranq : fig1:scattering ] . this measured partial current , normalized to the appropriate phase space element covered by the detector , yields the scattered current density . the double differential scattering cross - section , which is thus measured , is then defined by the ratio of this scattered current density to the incident neutron flux density , e.g. , @xmath40for each incident neutron in the plane wave state @xmath41 , the incident flux density is @xmath42 . the scattered current density is determined by the transition rate @xmath43 from the initial state @xmath44 , where the neutron is in the plane wave state @xmath45 with the spin @xmath46 and the scattering system is described by the set of variables @xmath47 , to the final state , @xmath48 . according to scattering theory @xcite , the transition rate is determined by the matrix elements of the transition operator ( or @xmath49-matrix ) @xmath50 , satisfying certain operator equations , which depend on the scattering system s hamiltonian , @xmath51 , and its interaction with the neutron , @xmath52 , @xmath53here @xmath54 is the scattering system s energy gain . it is convenient to introduce the scattering length operator , @xmath55 , which conveniently absorbs several factors , @xmath56 and its fourier transform , @xmath57 , @xmath58summing over all possible final scattering states , we obtain the double differential scattering cross - section for a given initial state , @xmath44 , @xmath59where the dependence on the spin - state of the neutron is implicit in @xmath60 . the energy and momentum transfer to the sample are governed by the conservation laws , @xmath61finally , following van hove @xcite , one can use the integral representation of the delta - function expressing the energy conservation in eq . ( [ tranq : eq : crosssec-2 ] ) , and the time - dependent scattering length operator whose evolution is governed by the system s hamiltonian , @xmath62to recast the double differential scattering cross - section in the most useful form of the two - time correlation function , @xmath63here the sum is over all possible final spin states of the scattered neutron , @xmath64 , since in the general case the scattering length operator , @xmath65 , depends on the neutron spin , @xmath66 . the sum over the final states of the sample has been absorbed into the expectation value of the two - time correlation function of the scattering length operator . the minus sign in front of @xmath67 in eq . ( [ tranq : eq : crosssec-3 ] ) follows from the convention adopted in the conservation laws in eq . ( [ tranq : eq : qe - conservation ] ) , where @xmath68 is the momentum transfer to the sample , which is the opposite of the change in the neutron s momentum . the total measured scattering cross - section is obtained by taking the proper thermal average of eq . ( [ tranq : eq : crosssec-3 ] ) over all possible initial states , @xmath69 . while the scattered neutron s wave vector @xmath39 is uniquely determined by @xmath35 and @xmath67 , by virtue of the conservation laws ( [ tranq : eq : qe - conservation ] ) , the neutron s spin state can be changed by transferring the angular momentum @xmath70 to the sample . in a polarized neutron experiment scattering between different neutron spin states can be measured . in such a case , the scattering length operator in eq . ( [ tranq : eq : crosssec-3 ] ) is a matrix with respect to different initial and final spin state indices ; it determines the various spin - flip and non - spin - flip cross - sections @xcite . in the more common case of unpolarized neutron scattering , neutron spin indices should be traced out in eq . ( [ tranq : eq : crosssec-3 ] ) , so that it determines a single unpolarized neutron scattering cross - section . finally , we should mention that the double differential cross - sections in eq . ( [ tranq : eq : crosssec-2 ] ) , ( [ tranq : eq : crosssec-3 ] ) are general expressions obtained from scattering theory and are valid for scattering of any probe particles . the remarkable advantage of neutron scattering is in the fact that scattering length operators are rather simple , very well understood , and are directly related to the fundamental physical properties of the scattering sample . he . the points show the compilation of the experimental neutron data presented in ref . the solid line is the fit of the low-@xmath71 part of the spectrum to the bogolyubov quasiparticle dispersion.,scaledwidth=75.0% ] for scattering from an individual nucleus , the scattering length operator can be very accurately approximated by a delta - function , eq . ( [ tranq : eq:2 - 5 ] ) . for a collection of nuclei in a condensed matter system , the total scattering length operator is obtained by adding scattering lengths of all nuclei , @xmath72where @xmath73 indexes the nucleus at position @xmath74 with scattering length @xmath75 . for a system of identical nuclei , this is just a particle number density operator in the scattering system , times @xmath25 , @xmath76substituting this into eq . ( [ tranq : eq : crosssec-3 ] ) and summing out the neutron s spin states we obtain , @xmath77therefore , the nuclear cross - section measures the space - time correlation of the atom number density in a condensed matter system . this is exactly the quantity of interest in many theories of strongly - correlated quantum systems . one of the first successes of neutron scattering was the measurement of the phonon - roton dispersion of the elementary excitations in superfluid helium-4 . neutron data have confirmed that the shape of the dispersion is in agreement with that previously postulated by landau and feynman , as illustrated in fig . [ tranq : fig : heliumdisp ] . this led to the broad acceptance of the neutron scattering technique as a prime tool for studying quantum systems . next we consider the case in which two or more types of nuclear scatterers ( with distinct scattering lengths @xmath78 and frequency of occurrence @xmath79 ) are present in the sample in a random fashion . for example , an element may have multiple isotopes , each with a distinct @xmath80 , or the nuclei have a spin , so that @xmath80 depends on the nuclear and neutron spin orientations , or we have at least two elements that are randomly distributed among equivalent positions . the average product of the scattering lengths for any two sites can then be written as @xmath81where @xmath82 we can then distinguish between coherent scattering , @xmath83which probes the inter - nuclear correlation , and the incoherent scattering , @xmath84which probes the local autocorrelation of the nuclear position ; the angle brackets denote the average over the sample state . in eqs . ( [ tranq : eq : crosssec - n - coh ] ) and ( [ tranq : eq : crosssec - n - incoh ] ) we have switched to the co - ordinate representation of nuclear density operator ( [ tranq : eq : b - nuclei - n ] ) and performed the fourier integration . as a result , nuclear positions @xmath74 and @xmath85 are quantum - mechanical operators and have to be treated appropriately in calculating the cross - section @xcite . in a crystal , the equilibrium positions of atomic nuclei are arranged on the sites of a lattice , so that the position of each individual nucleus @xmath73 can be represented as @xmath86where @xmath87 is the lattice site position , and @xmath88 is a small displacement of the atomic nucleus from its equilibrium position at @xmath87 . substituting this into eq . ( [ tranq : eq : crosssec - n - coh ] ) , one can show that the coherent nuclear cross - section of a monoatomic crystal is given by @xmath89here @xmath90 is the time- or lattice - averaged square of the atomic displacement from equilibrium in the direction of @xmath67 , and we have taken advantage of the fact that the correlation function in eq . ( [ tranq : eq : crosssec - n - coh ] ) depends only on relative coordinates , which allows one summation over the @xmath91 lattice sites to be completed . the integral contains an exponentiated correlation function of atomic displacements . it is useful to consider the series expansion of this term in powers of pair displacement correlations . in zeroth order , the exponential factor is just 1 , and one obtains the expression for the elastic bragg scattering in a crystal , @xmath92where we used the conventional notation for the debye - waller factor , with @xmath93 . using the lattice fourier representation , this can be recast in the common form @xmath94where @xmath95 is the reciprocal unit cell s volume ( @xmath96 is the volume of the unit cell in real space ) , and @xmath97 are the vectors of the reciprocal lattice . in a non - bravais crystal , where the unit cell contains several atoms , the sum in eq . ( [ tranq : eq : crosssec - n - coh - cryst ] ) has to be split into the intra - unit cell and the inter - unit cell parts , leading to @xmath98where the intra - unit cell summation yields the nuclear unit cell structure factor , @xmath99and @xmath100 indexes atoms in the unit cell . for some reciprocal lattice points , @xmath101 can be zero , which gives the bragg peak extinction rules in a non - bravais crystal . expanding the exponent in eq . ( [ tranq : eq : crosssec - n - coh - cryst ] ) to the first order , we obtain a contribution to the cross - section that is proportional to the correlation of displacements at two different sites . in cases where static disorder is present in the crystal , such as dislocations or lattice strain , the time - independent correlations of displacements between different sites give rise to elastic diffuse scattering . calculation of the time - dependent displacements of atomic nuclei from their equilibrium positions in the lattice is achieved by quantizing their vibrations in terms of quantum oscillators , called phonons . a phonon is a normal mode of atomic vibration , a coherent wave of atomic displacements in the crystal . we distinguish phonons with index @xmath102 . the polarization vector @xmath103 ( direction of atomic displacements ) and the dependence of the energy on the wave vector , @xmath104 ( dispersion ) , are determined by the the local inter - atomic potentials . the total number of such modes depends on the number of atoms in the unit cell of the crystal . only three phonons , which are all acoustic , are present for the bravais lattice , two transverse and one longitudinal . taking the proper thermal average over the sample s equilibrium state , the contribution to the neutron scattering is given by @xmath105 \ts , \label{tranq : eq : crosssec - phonon - brav}\index{phonons , neutron scattering cross section}\end{aligned}\]]where @xmath106 is the mass of each nucleus . the thermal factor @xmath107 is the bose distribution function describing thermal population of the oscillator states for temperature @xmath49 of the sample . the first term arises from phonon creation and corresponds to the neutron energy loss , while the second term is from an annihilation of a phonon that has been thermally excited in the crystal and results in the neutron energy gain . for a non - bravais crystal lattice , there are also optic phonons , arising from the different intra - unit - cell vibrations . the total number of phonons is equal to @xmath108 , the number of vibrational degrees of freedom of the @xmath109 atoms comprising the basis of the unit cell of the lattice . the contribution of each of these phonons to the neutron scattering cross - section is @xmath110 \ts , \label{tranq : eq : crosssec - phonon - nonbrav}\end{aligned}\]]where @xmath111 is the polarization vector for site @xmath100 in mode @xmath102 . for an acoustic phonon in the hydrodynamic , long - wavelength ( small @xmath71 ) and low - energy limit , this reduces to eq . ( [ tranq : eq : crosssec - phonon - brav ] ) , where the total mass of all atoms in the unit cell , @xmath112 , should be used and one must multiply by the square of the elastic structure factor , @xmath113 . the magnetic interaction of a neutron with a single atom is very weak , so the born approximation provides an extremely accurate account for magnetic neutron scattering by the atomic electrons . in this approximation , the transition matrix is given simply by the interaction potential , @xmath114 , where we have to combine the neutron s interaction with the electron s spin and orbital magnetic moment , eqs . ( [ tranq : eq : v - spin ] ) and ( [ tranq : eq : v - orb ] ) . accurate accounting for the orbital contribution to magnetic scattering presents , in general , a rather difficult and cumbersome task @xcite . there are some important cases where the orbital contribution is not significant , such as transition - metal atoms in a crystal , where the local crystal electric field typically quenches the orbital angular momentum , or the case of @xmath102-electrons , where @xmath115 . nevertheless , under some very general assumptions , the neutron s interaction with the electron orbital currents can be recast in the same way as its interaction with the spin magnetic moment , yielding for the total magnetic scattering length , @xmath116 \right ] \right ) \ts , \label{tranq : eq : b - mag}\index{magnetic scattering length}\end{aligned}\]]where @xmath117 is the sum of the spin and the orbital magnetization associated with each electron , @xmath118 . the fourier transform of the magnetic scattering length ( [ tranq : eq : b - mag ] ) , which determines the scattering cross - section , is @xmath119 \right ] \right ) \ts . \label{tranq : eq : b - mag - q}\]]here @xmath120 is the fourier transform of the total magnetization density of the atom , @xmath121 @xmath122 is the spin operator of @xmath123th electron , @xmath124 its orbital magnetic moment operator . the cross product in eq . ( [ tranq : eq : b - mag - q ] ) ensures the important property that only magnetization perpendicular to the wave vector transfer , @xmath67 , contributes to the magnetic neutron scattering . adding the contributions from all atoms in the crystal and averaging over the neutron polarizations , we obtain the magnetic neutron scattering cross - section measured in an experiment with unpolarized neutrons ( @xmath125 ) , @xmath126here @xmath127 is the fourier transformed magnetization density operator in the crystal . hence , magnetic neutron scattering measures the time- and space - dependent correlations of the magnetization fluctuations in the sample . introducing the dynamic correlation function , @xmath128we can rewrite eq . ( [ tranq : eq : crosssec - mag ] ) as @xmath129where @xmath130 cm is the characteristic magnetic scattering length . the dynamic correlation function defined above by eq . ( [ tranq : eq : smm - corr ] ) obeys two important relations that are derived in the linear response theory @xcite . first , it is the detailed balance constraint , which relates the energy gain and the energy loss scattering at a temperature @xmath49 , @xmath131the second is the fluctuation - dissipation theorem ( fdt ) , which relates the scattering intensity with the imaginary part of the dynamic magnetic susceptibility , @xmath132here @xmath133 and @xmath134 denote @xmath135 and @xmath136 symmetrized with respect to @xmath137 . a system with a center of inversion has symmetry with respect to @xmath138 , in which case the tildes can be dropped for the diagonal components in @xmath139 indices . this is the case for which the fdt is most frequently written @xcite . the fdt , eq . ( [ tranq : eq : fdt ] ) , is a consequence of the detailed balance condition ( [ tranq : eq : delbal ] ) and the causality relations , which require that @xmath140 is properly asymmetric . the fundamental laws of nature expressed in eqs . ( [ tranq : eq : delbal ] ) and ( [ tranq : eq : fdt ] ) are extremely useful in performing and analyzing neutron scattering experiments . if there exists a non - zero equilibrium magnetization in the sample , @xmath141 , where the bar over @xmath142 denotes the time - averaging , we can introduce magnetization fluctuation around this equilibrium , @xmath143 , and write @xmath144where the inelastic component @xmath145 is defined similarly to eq . ( [ tranq : eq : smm - corr ] ) , but with @xmath146 replaced by @xmath147 . the first term here leads to elastic scattering which results from static magnetization in the sample , while the second term describes the inelastic magnetic scattering arising from its motion . substituting the first term into eq . ( [ tranq : eq : crosssec - mag - s ] ) we obtain the unpolarized magnetic elastic cross - section , @xmath148where @xmath149 is the fourier transform of the magnetization component perpendicular to the wave vector transfer , @xmath67 . ( [ tranq : eq : crosssec - mag - el ] ) applies equally well to all cases where static magnetism is present in a crystal , whether it is a long - range magnetic order leading to bragg peaks , or a short - range , e. g. nano - scale magnetic correlation , resulting in an appearance of a broad magnetic diffuse scattering . in the case of a long - range order , the magnetization density in a crystal typically has an equilibrium static component , which is modulated with a wave vector @xmath150 , @xmath151where @xmath152 is a real vector function that describes the ferromagnetic component , if present , while @xmath153 can be complex and describes the staggered magnetization . these `` bloch amplitudes '' are periodic in the crystal lattice , and therefore can be expanded in the fourier series , @xmath154where the integral is over the unit cell of the nuclear ( paramagnetic ) crystal lattice . substituting ( [ tranq : eq : m_r ] ) and ( [ tranq : eq : m_tau ] ) into eq . ( [ tranq : eq : crosssec - mag - el ] ) , we obtain the following expression for magnetic bragg scattering associated with the long - range magnetic order at a wave vector @xmath150 , @xmath155 \right ) \delta(\hbar\omega ) \ts . \label{tranq : eq : crosssec - mag - el2}\index{bragg diffraction , magnetic}\end{aligned}\]]here the summation is over the paramagnetic crystal lattice . this is the `` large brillouin zone '' description , which is the most general one , in that it does not rely on the existence of a commensurate magnetic superlattice with a unit cell containing some integer number of nuclear lattice unit cells , and applies to incommensurate , as well as commensurate magnetic structures . such a description is most convenient for stripe phases in the cuprates , which are often incommensurate . the intensities of magnetic satellites , @xmath156 , are given by the fourier amplitudes of the magnetization , ( [ tranq : eq : m_tau ] ) , which are obtained by performing the fourier integrals over the unit cell of the paramagnetic lattice . in the case where the unit cell magnetization could be approximated by a number of point - like magnetic dipoles @xmath157 located at positions @xmath158 , these amplitudes become the conventional unit cell magnetic structure factors , @xmath159 in discussing magnetic scattering we assume a rigid lattice , neglecting atomic displacements due to disorder and vibrations discussed above . the leading correction to this description is obtained by multiplying expressions for magnetic cross - section with the debye - waller factor , @xmath160 . in many important cases the magnetization density in the crystal is carried by electrons localized on atomic - like orbitals , which are specified by the local atomic variables , such as spin and orbital quantum numbers . in such cases , the matrix element of the atomic magnetization in the magnetic neutron scattering cross - section can be factorized into the product of the reduced matrix element ( form factor ) , which does not depend on the direction of the atom s angular momentum quantum numbers , and the wigner 3j - symbol , which entirely accounts for such dependence . hence , the cross - section can be expressed in terms of a product of the @xmath161dependent form factor , which accounts for the shape of the magnetization cloud associated with the atomic spin and orbital variables , and a dynamical correlation function between these local angular momentum variables at different lattice sites . for magnetic ions obeying hund s rule , neutron scattering usually probes states belonging to the same multiplets of the angular momentum , @xmath162 , @xmath163 for the russel - saunders atoms with weak spin - orbit and strong crystal field , or @xmath164 for the case of strong spin - orbit coupling , where the total angular momentum @xmath165 is a good quantum number , such as in rare earths . hence , we can write for the fourier transform of atomic magnetization , @xmath166where the spin and the orbital magnetic form factors are , @xmath167@xmath168where we made explicit that initial and final states of the sample belong to the same @xmath169 and @xmath170 multiplet . similar relations hold for the @xmath171 multiplet in the strong spin - orbit coupling limit . typically it is possible to define an effective spin operator , @xmath172@xmath173where @xmath16 and @xmath174 are the effective @xmath175factors , @xmath176 , @xmath177 . these expressions are exact in the cases of a @xmath171-multiplet , where @xmath178 , @xmath179 and @xmath180 , or a pure spin multiplet , where @xmath181 and the orbital contribution is absent . they give the leading - order approximation in other cases . if the orbital moment is nearly quenched , as it is for magnetic @xmath182-elements in strong crystal field , then @xmath183 , @xmath184 , and the orbital contribution to @xmath185 , is small . assuming this to be the case , we shall omit tildes and use @xmath186 for the effective spin . using the factorization of atomic magnetization provided by eqs . ( [ tranq : eq : s_we ] ) and ( [ tranq : eq : seff_ff ] ) , the magnetic neutron scattering cross - section ( [ tranq : eq : crosssec - mag ] ) can be recast as @xmath187where we allow for the possibility that the @xmath175factor is anisotropic , and that both @xmath188 and @xmath189 could be different for different sites @xmath190 of the lattice . equation ( [ tranq : eq : crosssec - spin ] ) relates the magnetic cross - section to the dynamic spin structure factor , which is the fourier transform of the time - dependent two - point correlation function of the atomic spin variables on the sites of the lattice , @xmath191@xmath192 is a quantity which is calculated in theoretical models based on the local spin hamiltonians . it also obeys a number of important relations , known as sum rules , which are extremely useful in analyzing neutron scattering data . the zero moment sum rule is obtained by integrating eq . ( [ tranq : eq : s - ss - corr ] ) in @xmath67 and @xmath193 , providing the direct connection of the integral neutron intensity with the spin value @xmath170 in the lattice spin hamiltonian , @xmath194the first moment sum rule relates @xmath195 , which is the integral oscillator strength of the fluctuation spectrum , with the bond energies in the spin hamiltonian , and so on . representing the neutron scattering cross - section via two - point dynamical spin correlation function , as in eq . ( [ tranq : eq : crosssec - spin ] ) , is possible in a large number of important magnetic systems , such as cuprates and other @xmath196 magnetic insulators . such a representation is extremely useful , as it allows one to connect the measured magnetic neutron intensity with the theoretically predicted properties of model spin hamiltonians , such as the heisenberg spin hamiltonian , @xmath197 here @xmath198 is the exchange coupling between sites @xmath73 and @xmath199 , and @xmath200 and @xmath201 are the lattice fourier transforms , @xmath202 in many systems with magnetic order , the average value of spin at each lattice site in the ground state ( gs ) is `` frozen '' at nearly the full saturation value , @xmath203 . in particular , this is a very good approximation for the semi - classical spins , @xmath204 , in more than one dimension ( 1d ) . for quantum spins , @xmath205 , and/or in the low - dimensional , or frustrated systems , the order may be weak , or absent , and such a picture is inadequate . nevertheless , in a large number of systems magnetic order in the ground state is well developed , and the semiclassical spin - wave picture applies . spin excitations in a magnetic system with a well - ordered ground state , such as a ferromagnet , where all spins are parallel , or a semi - classical antiferromagnet , where there are two antiparallel sublattices , can be visualized as small oscillations of classical spin vectors around their equilibrium positions in the gs spin structure . their wave - like spatial composition results from the translational symmetry of the system . frequencies of such spin - wave oscillations can be calculated from the spin hamiltonian , such as ( [ tranq : eq : heis_spin_hamilt ] ) , totally within the classical mechanics , simply by writing the torque equations of motion for the classical spin angular momenta . for example , in the case of the heisenberg hamiltonian ( [ tranq : eq : heis_spin_hamilt ] ) for a magnetically ordered system characterized by the ordering wave vector @xmath206 ( this includes ferromagnetism with @xmath207 , as well as antiferromagnetism and helimagnetism ) , one obtains the spin - wave dispersion @xcite , @xmath208this can be recast as @xmath209 , where @xmath210 . spin waves are the normal modes of the linearized equations of motion . they involve small spin deviations that are perpendicular to the equilibrium spin direction . hence , spin waves are transversely polarized , with two mutually orthogonal linear polarizations of spin oscillations possible . for a spin system on a bravais lattice there are two spin - wave modes . in a quantum - mechanical treatment of spins , the spin - wave calculation proceeds via an approximate mapping of spin operators to bose creation - annihilation operators , _ i.e. _ to local oscillator modes . hence , the so obtained spin - wave theory ( swt ) describes spin excitations as coherent waves of small oscillations around the local equilibrium positions , in many ways similar to phonons . the resulting expression for the spin - wave contribution to the neutron magnetic scattering cross - section in a sample with a spiral spin structure with the propagation vector @xmath206 is @xmath211 , \;\;\ ; \label{tranq : eq : crosssec - sw - spiral}\index{spin waves , neutron scattering cross section}\end{aligned}\]]where @xmath212 is the direction normal to the plane of the spiral , and we have restricted consideration to the case of a bravais lattice and retained only the contribution corresponding to creation of a single spin wave . the contribution arising from the absorption of a spin wave is written similarly to that of a phonon in eq . ( [ tranq : eq : crosssec - phonon - brav ] ) . for a ferromagnet , the single spin - wave magnetic cross - section simplifies to , @xcite , @xmath213 , \;\;\ ; \label{tranq : eq : crosssec - sw - fm}\index{spin waves , neutron scattering cross section}\end{aligned}\]]where @xmath214 is the wave vector component along the ferromagnetic ordered moment and we have retained the contributions from both the creation and the absorption of a spin wave . it is clear from eq . ( [ tranq : eq : crosssec - spin ] ) that even the exact knowledge of the dynamical spin structure factor ( available from theory in some special cases , such as in one dimension ) is insufficient to reproduce the measured magnetic scattering cross - section . one also has to know the magnetic form factor , which needs to be obtained from an ab initio calculation of the electronic density in the crystal . in the most common case of a hund s ion with @xmath215 unpaired electrons forming spin @xmath216-multiplet , the spin magnetic form factor ( [ tranq : eq : s_ff ] ) becomes @xmath217where the sum is only over the unpaired electrons . the single - electron density , @xmath218 , is determined from the many - electron atomic wave function through @xmath219 . the magnetic form factor for an atom is therefore simply an average of those for each of the unpaired electrons . similarly , the orbital form factor is the fourier - transformed average density of the uncompensated orbital currents in the atom . if the average hartree - fock potential acting on an unpaired electron @xmath118 in the atom is spherically symmetric , then the effective one - electron wave functions in ( [ tranq : eq : s_ff_hund ] ) are the eigenfunctions of angular momentum and are tagged by the @xmath220 quantum numbers , @xmath221 . the angular and the radial dependencies of the electronic density factorize , @xmath222 , where @xmath223 is the spherical function giving the dependence on the polar angles @xmath224 . this so - called central field approximation is good when the contribution to the potential from electrons in the incomplete shell is small . however , it also becomes exact for an almost - filled shell , with only a single electron , or a single hole , as in the case of cu@xmath225 , or for a nearly half - filled shell , because the average potential of the closed , or half - filled shell , is spherically symmetric . in the general case , a single - electron wave function can always be expanded in a series in spherical harmonics . in each term of such an expansion , the radial and the angular parts are again factorized , and the magnetic form factor is a sum of fourier - transformed terms with different @xmath226 and @xmath227 . the same kind of an expansion is encountered in calculating the orbital contribution to the magnetic form factor . this is known as a multipole expansion @xcite . the calculations are ion - specific and extremely cumbersome . the general expressions can be obtained only for the leading , isotropic contributions , in the limit of small wave vector transfer , known as the dipole approximation , @xmath228where @xmath229 and @xmath230 are the @xmath231 dependent radial integrals quantifying the radial wave function @xcite . the radial integrals for most known magnetic atoms and ions have been calculated numerically from the appropriate hartree - fock or fock - dirac wave functions and are tabulated in ref @xcite . the full @xmath232 is given by eq . ( [ tranq : eq : seff_ff ] ) . although the dipole approximation ( [ tranq : eq : s_ff_dipole ] ) is the one most commonly used , it is extremely crude . in particular , it does not account for the anisotropy of the magnetic form factors , which can be very important for ions with only one or two unpaired electrons . the anisotropic magnetic form factor of a single @xmath233 hole in a @xmath234 orbital of the magnetic ir@xmath235 ion in the cubic k@xmath13ircl@xmath236 was studied in ref . the authors found that the anisotropy of the magnetic form factor is very large , with an additional enhancement coming from the hybridization of the ir @xmath233-orbital with the cl @xmath237-orbitals . the anisotropy of the magnetic form factor is also very pronounced in la@xmath13cuo@xmath2 , yba@xmath13cu@xmath238o@xmath239 , and related cuprate materials , including the high-@xmath240 superconductors , where in the ionic picture a single unpaired magnetic electron occupies a @xmath241 orbital . in ref . @xcite the authors found that properly accounting for the anisotropy of the cu@xmath225 magnetic form factor is essential for understanding the magnetic bragg intensities measured in yba@xmath13cu@xmath238o@xmath239 at large wave vectors , and can also explain the peculiar @xmath67-dependence of the inelastic magnetic cross - section in this material . accounting for the anisotropic cu@xmath225 form factor was also very important in analyzing neutron scattering by high - energy spin waves in la@xmath13cuo@xmath2 @xcite , and the chain cuprates srcuo@xmath13 and sr@xmath13cuo@xmath238 @xcite . the magnetic excitations in these cuprate materials extend to several hundreds of mev . consequently , the measurements require very large wave vector transfers , for which the anisotropy of the magnetic form factor is very pronounced . the ionic magnetic form factors for @xmath196 orbitals can be explicitly computed by fourier transforming the corresponding spherical harmonics . in particular , for the @xmath242 orbital relevant for cu@xmath225 one obtains @xcite , @xmath243where @xmath244 are the polar angles of the wave vector @xmath67 in the local coordinate system used to specify the proper orbital wave functions in the crystal field . although using the anisotropic ionic magnetic form factor of cu@xmath225 is much better than using a spherical form factor of the dipole approximation , it is still not sufficient for cuprates , as it neglects the effects of covalency ( i.e. , charge transfer to the neighboring oxygen ) that are expected to be very significant in these materials . in ref . @xcite it was discovered that covalent bonding results in a marked modification of the magnetic form factor in the quasi-1d antiferromagnet sr@xmath13cuo@xmath238 . the local structure of the planar cu o square plaquettes in this material is essentially identical to that in la@xmath13cuo@xmath2 . making use of a precise theoretical result for the excitation spectrum available in 1d , the authors demonstrated that a good fit to the data requires a form factor that takes account of hybridization between the half - filled cu @xmath241 orbital and the ligand o @xmath245 orbitals , as given by a density functional calculation . the hybridization causes the spin density to be extended in real space , resulting in a more rapid fall off in reciprocal space compared to a simple cu@xmath225 form factor , as illustrated in fig . [ tranq : fig : ff2_cu ] . smaller values of magnetic form factor at relatively large wave vectors , where the measurement is performed , lead to the suppression of magnetic intensity , which could be as large as a factor of two or more @xcite . finally , we note that a study of covalent nmr shifts by walstedt and cheong @xcite found that barely 2/3 of the spin density in la@xmath13cuo@xmath2 resides on the copper sites , in excellent agreement with the sr@xmath13cuo@xmath238 neutron data of walters _ et al . _ @xcite . given by eq . ( [ tranq : eq : cu-3d - ff ] ) ( solid line ) and the covalent magnetic form factors for sr@xmath13cuo@xmath238 ( dashed line ) and la@xmath13cuo@xmath2 ( dash - dotted line ) obtained from the ab initio density functional calculations @xcite . panels ( a ) - ( c ) show the dependence along three principal directions . ] the discovery of high - temperature superconductivity in la@xmath0ba@xmath1cuo@xmath2 ( lbco ) came as a considerable surprise @xcite , as ceramic oxides were generally considered to be poor conductors . the structure of lbco and related cuprates involves cuo@xmath13 layers , with the cu atoms forming a square lattice with bridging o atoms , as shown in fig . [ tranq : fg : af](a ) . anderson @xcite predicted that the parent compound , la@xmath13cuo@xmath2 , should have strong antiferromagnetic ( af ) superexchange interactions between nearest - neighbor cu atoms . the occurrence of antiferromagnetic order was demonstrated by vaknin _ _ @xcite using neutron diffraction on a powder sample of la@xmath13cuo@xmath2 . as illustrated in fig . [ tranq : fg : af](b ) , the antiferromagnetic nel order doubles the size of the unit cell in real space , which results in magnetic superlattice peaks , as shown in ( c ) . thus , the antiferromagnetic order can be detected through the appearance of superlattice peaks . the challenge in this case is that one must distinguish from structural superlattice peaks due to staggered rotations of cuo@xmath236 octahedra @xcite . fortunately , the af and structural peaks appear at inequivalent positions . plane , with cu atoms indicated by filled circles and o atoms by open diamonds . ( b ) schematic of antiferromagnetic order , with alternating up ( filled circles ) and down ( open circles ) spins . the solid line indicates the chemical unit cell , while the dashed line indicates the doubled area of the antiferromagnetic unit cell . ( c ) reciprocal space showing fundamental bragg peak positions ( filled circles ) and antiferromagnetic superlattice peak ( open circle ) at @xmath246 . ] the ordered pattern of the octahedral tilts is associated with an orthorhombic distortion of the crystal structure that makes the diagonal directions of a cu - o plaquette inequivalent @xcite , as indicated in fig . [ tranq : fg : struc ] . by analyzing the @xmath67 dependence of the af bragg peak intensities , it was possible to determine that the magnetic moments on cu atoms lie within the cuo@xmath13 planes , pointing along the orthorhombic @xmath25 axis @xcite . furthermore , it was possible to show that the relative arrangement in neighboring planes is as shown in fig . [ tranq : fg : struc ] . with the magnetic structure determined , one can evaluate the magnitude of the magnetic moments by normalizing the af peak intensities to the nuclear intensities and correcting for the magnetic form factor . early studies yielded a small ordered moment whose magnitude was correlated with the magnetic ordering , or nel , temperature , @xmath247 @xcite . neutron scattering studies on carefully prepared single - crystal samples eventually demonstrated the impact of interstitial oxygen , within the la@xmath13o@xmath13 layers @xcite . removing the excess oxygen by annealing , one can achieve @xmath248 k @xcite and a magnetic moment of @xmath249 @xmath250 @xcite . cuo@xmath2 , with arrows indicating the arrangement of the magnetic moments in the antiferromagnetic state . figure reprinted with permission from lee _ et al . _ @xcite . copyright ( 1999 ) by the american physical society . ] if one assumes a @xmath16 factor of roughly 2 , then the ordered moment yields an average ordered spin @xmath251 , compared to the expected @xmath252 per cu atom . the reduction results from the strongly anisotropic structure and the low value of the spin . for a two - dimensional magnetic system described by a heisenberg spin hamiltonian , long - range order is destroyed at any finite temperature by thermal excitation of spin fluctuations . for la@xmath13cuo@xmath2 , weak ( nearly - frustrated ) couplings between the planes enable the ordering at finite temperature @xcite . nevertheless , the spin correlations have a strongly two - dimensional ( 2d ) character , as demonstrated by neutron scattering studies @xcite . the small magnitude of the spin , combined with the enhanced zero - point spin fluctuations in 2d , puts the system close to a quantum critical point @xcite . although the large fluctuations cause problems for perturbation theory , spin - wave theory nonetheless yields a result , @xmath253 , that is very close to the value obtained from experiment @xcite . the exchange couplings between the spins can be determined by analyzing the dispersion of the spin excitations , which can be obtained by inelastic scattering measurements on a single - crystal sample . early studies of la@xmath13cuo@xmath2 with triple - axis spectrometers demonstrated that the superexchange energy @xmath171 coupling nearest - neighbor spins is greater than 100 mev , and that effects such as exchange anisotropy and interlayer coupling are very small @xcite . time - of - flight techniques were required to measure the highest - energy spin waves @xcite , and these have been refined over time @xcite . the most recent results , from headings _ et al . _ @xcite , are shown in fig . [ tranq : fg : lco ] . the line through the data points corresponds to a fit with linear spin - wave theory , which works surprisingly well in light of the the large zero - point fluctuations . one impact of the latter is the renormalization factor , with a fitted value of @xmath254 , that is required to fit the measured intensity . this is somewhat smaller than the value @xmath255 that is predicted from quantum corrections to linear spin waves @xcite . it should be noted , though , that this analysis did not take account of the hybridization effects on the magnetic form factor , discussed in sec . [ tranq : sc : mff ] , which would account for some of the apparent renormalization . an additional effect is the damping and broadening of the energy - dependence of the spin - wave line shape at the zone boundary position @xmath256 . this appears to be the result of interactions with a multi - magnon high - energy continuum @xcite . cuo@xmath2 at @xmath257 k. lines through data correspond to fits with spin - wave theory ; the fit to the intensity includes a renormalization factor @xmath254 . figure reprinted with permission from headings _ et al . copyright ( 2010 ) by the american physical society.,scaledwidth=60.0% ] the fitted dispersion corresponds to @xmath258 mev , but also requires longer - range exchange couplings second and third neighbor couplings @xmath259 and @xmath260 , which are relatively weak , and a significant 4-spin cyclic exchange term @xmath261 . the overall bandwidth of the magnetic spectrum is @xmath262 . a recent analysis of the couplings , including @xmath263 , in terms of a single - band hubbard model has been given by dalla piazza _ _ @xcite . to achieve superconductivity , one must dope charge carriers into the cuo@xmath13 planes . substituting ba@xmath225 or sr@xmath225 for la@xmath264 introduces holes . a small density of holes , @xmath265% , is enough to kill the long - range af order , which is followed by a regime of spin - glass order @xcite . doping beyond @xmath266 yields superconductivity . the maximum superconducting transition temperature @xmath240 occurs for @xmath267 , with @xmath240 heading towards zero for @xmath268 . inelastic neutron scattering studies have been performed on single crystal samples across this entire doping range @xcite . a couple of the key features are : 1 ) the bandwidth of strong spin - fluctuation scattering decreases linearly with doping , being quantitatively similar to the pseudogap energy extracted from various electron spectroscopies @xcite , and 2 ) the wave vector characterizing the low - energy spin excitations splits about the af wave vector , becoming incommensurate @xcite . insight into the cause of the magnetic incommensurability was provided by neutron diffraction measurements on a closely related material , la@xmath269nd@xmath270sr@xmath271cuo@xmath2 @xcite . the impact of the nd substitution is to modify the tilt pattern of the cuo@xmath236 octahedra such that the in - plane cu - o bond directions become inequivalent @xcite . new superlattice peaks were observed in this low - temperature phase , with in - plane wave vectors @xmath272 and @xmath273 corresponding to spin order and @xmath274 and @xmath275 associated with modulations of atomic positions due to charge order , with @xmath276 . such results have been confirmed in the system la@xmath0ba@xmath1cuo@xmath2 @xcite . analysis of the superlattice peaks indicates that they are evidence for spin and charge stripe order @xcite , as illustrated in fig . [ tranq : fg : stripes ] . because of the crystal symmetry , the orientation of the stripes rotates @xmath277 from one layer to the next . plane ( only cu sites shown ) . note that the magnetic period is twice that of the charge period . the charge density along a stripe is one hole for every two sites in length . the registry of the stripes with respect to the lattice ( for example , site - centered vs. bond - centered ) has not yet been determined experimentally . ] the occurrence of maximum stripe order corresponds to a strong suppression of the bulk @xmath240 at @xmath278 @xcite , suggesting that stripe order competes with superconductivity ; however , recent studies have demonstrated that 2d superconductivity can coexist with stripe order @xcite . it now appears that superconducting order can intertwine with stripe order @xcite . thus , understanding stripe correlations may provide valuable insights into the nature of the superconducting mechanism of cuprates . neutron scattering on a time - of - flight instrument has been used to characterize the spin excitation spectrum in la@xmath0ba@xmath1cuo@xmath2 with @xmath279 @xcite . the effective dispersion and the @xmath67-integrated spectral weight are shown in fig . [ tranq : fg : disp ] . above 50 mev , the excitations disperse upwards like antiferromagnetic spin waves with an energy gap ; the solid line through the points in each panel corresponds to a two - leg spin ladder model with @xmath280 mev . below 50 mev , the excitations disperse downwards toward the positions of the incommensurate magnetic superlattice peaks . when the sample is warmed to a state with no static stripe order , the spectrum maintains its essential features @xcite . it appears that stripes , whether static or dynamic , provide a way for the superexchange mechanism to survive when the antiferromagnetic layers are doped with holes . -integrated spectral weight and ( b ) effective magnetic dispersion in the stripe - ordered phase of la@xmath0ba@xmath1cuo@xmath2 with @xmath279 , from @xcite . the solid lines through the data points are described in the text . in ( a ) , the peak at @xmath281 mev is now know to be due to a phonon mode . in ( b ) , the effective dispersion is plotted for @xmath282 along a line through the incommensurate magnetic superlattice peaks . ] the relevance of charge - stripe order is less clear in cuprates families such as yba@xmath13cu@xmath238o@xmath239 and bi@xmath13sr@xmath13cacu@xmath13o@xmath283 ; nevertheless , the dispersion of the magnetic excitations in these compounds ( measured by neutron scattering ) has been shown to be quite similar to that of lbco @xcite . the main difference is that the low - energy excitations tend to be gapped in the superconducting state , with a pile up of weight ( `` resonance '' peak ) appearing above the gap for @xmath284 . the commonality of the dispersions over a broad energy range suggests that the charge and spin correlations in superconducting and striped cuprates are similar . m. fujita , h. hiraka , m. matsuda , m. matsuura , j.m . tranquada , s. wakimoto , g. xu , k. yamada , j. phys . jpn . * 81 * , 011007 ( 2012 ) . p. bourges , y. sidis , c. r. phys . * 12 * , 461 ( 2011 ) . mohr , b.n . taylor , d.b . newell . . e. fermi , l. marshall , phys . rev . * 71 * , 666 ( 1947 ) . e. fermi , l. marshall , phys . rev . * 72 * , 408 ( 1947 ) . a.j.c . wilson ( ed . ) , _ international tables for crystallography _ , vol . c ( kluwer academic publishers , dordrecht , 1995 ) b.a . lippmann , j. schwinger , phys . rev . * 79 * , 469 ( 1950 ) . l. van hove , phys . rev . * 95 * , 1374 ( 1954 ) . m. blume , phys . * 124 * , 96 ( 1961 ) . donnelly , j.a . donnelly , r.n . hills , j. low temp . phys . * 44 * , 471 ( 1981 ) j. villain , j. phys . solids * 11 * , 303 ( 1959 ) . a. yoshimori , j. phys . soc . jpn . * 14 * , 807 ( 1959 ) . lynn , g. shirane , m. blume , phys . * 37 * , 154 ( 1976 ) . s. shamoto , m. sato , j.m . tranquada , b.j . sternlieb , g. shirane , phys . b * 48 * , 13817 ( 1993 )	neutron scattering is a powerful probe of strongly correlated systems . it can directly detect common phenomena such as magnetic order , and can be used to determine the coupling between magnetic moments through measurements of the spin - wave dispersions . in the absence of magnetic order , one can detect diffuse scattering and dynamic correlations . neutrons are also sensitive to the arrangement of atoms in a solid ( crystal structure ) and lattice dynamics ( phonons ) . in this chapter , we provide an introduction to neutrons and neutron sources . the neutron scattering cross section is described and formulas are given for nuclear diffraction , phonon scattering , magnetic diffraction , and magnon scattering . as an experimental example , we describe measurements of antiferromagnetic order , spin dynamics , and their evolution in the la@xmath0ba@xmath1cuo@xmath2 family of high - temperature superconductors .
recent years witnessed large developments in the kinetic theory methods applied to mathematical physics and more recently also to mathematical biology . among important branches of the kinetic theory are optimal transportation problems and related to them wasserstein metrics , and monge - kantorovich metrics @xcite . partial differential equations in metric spaces are being applied to transportation problems @xcite , gradient flows @xcite and structured population models @xcite . output of mathematical modeling can often be described as radon measures . comparing the results of the models requires then a definition of distance in the space of measures . the desired properties of such metrics depend on the structure of the considered problem . in most of the cases , the topology of total variation is too strong for applications , and weaker metrics had to be introduced , see @xcite for details . well known @xmath2-wasserstein metric , on the other hand , is only applicable to processes with mass conservation . to cope with growth in the process , various modifications have been proposed , including flat metric and centralized wasserstein metric . in the present paper , we additionally introduce a normalized wasserstein distance . for comparison of different metrics , their interpretation and examples see table [ table ] . even though , all those distances can be computed using linear programming ( lp ) , its computational complexity becomes often larger than the complexity of solving the original problem . for example , the equations for which the stability of numerical algorithm is proven in @xmath3 require an efficient algorithm for the flat metric to find the residual error . algorithms proposed in this paper are designed to compute efficiently wasserstein - type distance between two radon measures in the form of @xmath4 , where @xmath5 is the dirac delta . the algorithms can be also applied for the case of an arbitrary pair of measures , by approximating those measures by the sum of dirac deltas . in the process of finding numerical solutions to partial differential equations , the distance between a discrete and absolutely continuous measures is needed to evaluate the quality of the initial condition approximation , while the distance between two discrete measure is needed to find the residual error . the paper is organized as follows . in section [ metrics ] we introduce and compare four wasserstein - type metrics . additionally , we introduce a two - argument function , based on the wasserstein distance , which is not a metric , however , provides a good tool to estimate the flat metric from above . section [ algorithms ] is devoted to numerical algorithms proposed to calculate distances between two measures in respect to the considered metrics . we justify the algorithms and provide respective pseudocodes . the novelty of this paper is the algorithm for flat metric , which has been recently proven to be adequate for the escalator boxcar train ( ebt ) method for solving transport equations @xcite . we propose an efficient algorithm which computation cost is @xmath6 and optimize it further to @xmath1 . to judge the efficiency of the algorithms , we compare the times needed to compute the flat distance between two sums of a large number of dirac deltas randomly distributed over [ -1,1 ] . the framework of wasserstein metric in the spaces of probability measures has proven to be very useful for the analysis of equations given in a conservative form , such as , for example , transport equation @xcite , fokker - planck @xcite or nonlinear diffusion equation @xcite . it was originally defined using the notion of optimal transportation , but we focus on its dual representation . @xmath2-wasserstein distance between two measures @xmath7 and @xmath8 is given by @xmath9 observe that the integral @xmath10 is invariant with respect to adding a constant to function @xmath11 . consequently , for any @xmath12 holds @xmath13 furthermore , the following two properties hold : * @xmath2-wasserstein distance is scale - invariant @xmath14 * @xmath2-wasserstein distance is translation - invariant @xmath15 if @xmath16 , then @xmath17 by the dual representation definition . it makes this metric useless for processes not preserving the conservation of mass . in this section we introduce a simple normalization that leads to a definition of translation - invariant metric suitable for non - local models we define normalized @xmath2-wasserstein distance between two measures @xmath7 and @xmath8 as @xmath18 where @xmath19 denotes the total variation of the measure @xmath7 . the distance defined by is a metric . let @xmath7 , @xmath8 and @xmath20 be radon measures . then , it holds * @xmath21 if and only if @xmath22 . + indeed , either @xmath23 or @xmath24 imply that @xmath22 . * @xmath25 , * since @xmath26 to show the triangle inequality , we consider four possibilities @xmath27 this metric can be easily computed numerically using the algorithm for @xmath2-wasserstein distance . however , it lacks the scaling property , which holds for the wasserstein distance , and which is useful for applications . nonetheless , the following weaker property holds . let @xmath28 and @xmath29 be two sequences of radon measures and @xmath30 , @xmath31 then @xmath32 in this section we present a different modification of the @xmath2-wasserstein distance , which is scale - invariant . we define centralized @xmath2-wasserstein distance between two measures @xmath7 and @xmath8 as @xmath33\right\}.\ ] ] the metric was introduced in @xcite for analysis of the structured population models in the spaces of non - negative radon measures , @xmath34 , on @xmath35 with integrable first moment , i.e. , @xmath36 this metric satisfies the scaling property , but on the contrary to wasserstein metric , it is not invariant with respect to translations . it is therefore only useful for modeling specific phenomena , for example structured population dynamics where mass generation can only occur in one specific point of space . another solution of the problem of comparing two measures of different masses is the flat metric which is defined as follows flat distance between two measures @xmath7 and @xmath8 is given by @xmath37 the flat metric , known also as a bounded lipschitz distance @xcite , corresponds to the dual norm of @xmath38 , since the test functions used in the above definition are dense in @xmath39 . this metric satisfies the scaling property and it is translation invariant . it has proven to be useful in analysis of structured population models and in particularly , lipschitz dependence of solutions on the model parameters and initial data @xcite . the flat metric has been recently used for the proof of convergence and stability of ebt , which is a numerical algorithm based on particle method , for the transport equation with growth terms @xcite . consequently , an implementation of the ebt method requires an algorithm to compute the flat distance between two measures . computing of flat metric exactly is more costly than computing of @xmath2-wasserstein metric ( @xmath1 vs @xmath40 , see section [ sectionflatalg ] ) . moreover , often in applications it is sufficient to calculate upper bound of the distance , for example when computing residual error . we propose the following function , which requires only linear computing time . @xmath41 let @xmath7 , @xmath8 be radon measures on @xmath0 . then , the following estimate holds @xmath42 assume that @xmath43 . then , using the estimate of the flat metric by @xmath2-wasserstein distance for the measures with equal masses provides @xmath44 the last equality results from the representation of radon distance and positivity of measure @xmath8 . the upper bound function is more useful if it can be estimated from above by @xmath45 for some constant @xmath46 . in a general case , however , such constant does not exist , since by taking @xmath47 , @xmath48 and passing @xmath49 we obtain @xmath50 nevertheless , the desired estimate can be shown on a bounded set . let @xmath7 , @xmath8 be non - negative radon measures on a compact set @xmath51 , then the following estimate holds @xmath52 let @xmath53 and @xmath54 be such point that @xmath55 for any @xmath56 . then , assuming @xmath43 , we obtain @xmath57 in the following table we compare the introduced metrics and provide examples and interpretations in terms of optimal transport . @xmath58 & scale - invariance & translation - invariance & intuition of @xmath59 & compute complexity + wasserstein & @xmath60 & yes & yes & the cost of optimal transportation of distribution @xmath7 to a state given by @xmath8 , assuming that moving mass @xmath61 by @xmath62 requires @xmath63 energy . & @xmath40 + wasserstein normalized & @xmath64 @xmath65 & & & minimum of the cost of annihilating @xmath7 and generating @xmath8 ; and of the cost of generating the difference in mass between @xmath7 and @xmath8 and transporting @xmath66 to @xmath67 , assuming that generating / annihilating mass at any position requires the cost equal to the mass . & @xmath40 + wasserstein centralized & @xmath68 & yes & no & the cost of generating the difference in mass in point @xmath69 in space added to the cost of transporting @xmath70 to @xmath8 . & @xmath40 + flat & @xmath71 & yes & yes & the cost of optimal transporting and/or generating and/or annihilating mass to form @xmath8 from @xmath7 . & @xmath1 + radon & @xmath72 & yes & yes & the cost of generating and/or annihilating mass to form @xmath8 from @xmath7 & @xmath40 + in the remainder of this paper , we present algorithms for computing and approximating of the introduced metrics . we start with introducing the algorithm for computation of @xmath2-wasserstein distance . the algorithm is well - known and straightforward . nevertheless , we present it here , since the specific approach used in this section will be also applied later in the proof of correctness of more involved algorithms for other distances . define @xmath77}{n}}}&=&\sum_{i=1}^{n}\delta_{i / n}\mu\left[\frac{i-1}{n},\frac{i}{n}\right).\\\end{aligned}\ ] ] then , we estimate @xmath78}{n } } ) \leq \sum_{i=1}^{n}\frac{1}{n } \mu\left[\frac{i-1}{n},\frac{i}{n}\right ) \leq \frac{1}{n } \mu[0,1],\ ] ] and the right - hand side tends to @xmath69 with @xmath79 . consequently , @xmath2-wasserstein distance between an arbitrary pair of finite radon measures can be estimated by the distances between their discrete approximations . from this point on , we assume that @xmath80 define @xmath85 then , obviously @xmath86 . denote @xmath87 , and observe that @xmath88 @xmath89}w_{1}^{1}(x)=a_{2}x+a_{1}x+a_{1}\cdot sgn(a_{1})d_{1}=\\ & = & \left(a_{1}+a_{2}\right)x+\|a_{1}\|d_{1}.\end{aligned}\ ] ] it can be shown by induction that @xmath90 notice that the value @xmath91 is not used in this formula . it is , however , involved indirectly , as @xmath92 . we assume @xmath108 define @xmath109 as already proven @xmath110 from lp representation of the metric @xmath111 it can be deduced that @xmath112}\left(\underline{w}_{1}^{m+1}(x)+\overline{w}_{1}^{m+1}(x)-a_{m+1}x\right),\ ] ] so the distance is given by the formula @xmath113 each iteration of each loop takes a constant time . the total number of iterations in all three loops is equal to @xmath131 . computational complexity of this algorithm is therefore @xmath102 , while the memory complexity is @xmath103 . the algorithm for flat distance requires storing the shape of functions analogous to @xmath132 as they get more complicated when @xmath61 increases . in section [ flat_alg ] we provide a recursive formula for the sequence of these functions . the pseudocode in section [ flat_pseudo ] implements the algorithm using an abstract data structure to store previously defined functions . the computational complexity depends on the choice of this structure . in further sections we provide two solutions that require @xmath6 and @xmath1 operations . define @xmath138 obviously , it holds by definition @xmath139}f^{n}(x).\ ] ] observe that @xmath140\cap[-1,1]}f^{1}(x ) = a_{2}x+\min(\|a_{1}\|,a_{1}x+\|a_{1}\|d_{1}),\\\\\\ f^{m}(x ) & = & a_{m}x+\sup_{f_{m-1}\in[x - d_{m-1},x+d_{m-1}]\cap[-1,1]}f^{m-1}(x).\end{aligned}\ ] ] computing of @xmath141 based on @xmath142 is more complex than in previous cases , as @xmath142 is not necessarily monotonic . @xmath146\cap[-1,1]}f^{n}(x).\ ] ] choose @xmath147 $ ] . then , there exist @xmath148,\ , y'\in b(y , d)\cap[-1,1]$ ] such that @xmath149 because @xmath143 is concave , it holds @xmath150 the last inequality follows from @xmath151 . it is now proven that @xmath152 is convex , as it is a sum of a linear function and a convex function @xmath153 . for each @xmath61 the function @xmath143 is piecewise linear on @xmath61 intervals and it holds for some point @xmath154 @xmath155}}\\ f^{m-1}(x_m ) & \mbox{on \ensuremath{[x_m - d_{m-1},x_m+d_{m-1}]}}\\ f^{m-1}(x - d_{m-1 } ) & \mbox{on \ensuremath{[x_m+d_{m-1},1 ] } } \end{cases}\ ] ] @xmath156 is a linear function , so it can be described by its values in @xmath157 . assume that @xmath143 can be described by at most @xmath158 points and is linear between those points . as @xmath143 is concave , there exists a point @xmath159 $ ] such that @xmath160 for every @xmath62 . the maximum of @xmath143 on an interval whose both ends are smaller than @xmath161 is taken on its right end . similarly , if both ends of the intervals are larger than @xmath161 , the maximum is taken on its left end . finally , if the interval contains @xmath161 , the maximum is in @xmath161 . these considerations prove the formula for @xmath162 . it follows that @xmath163 is piecewise linear and it can be described by as many points as @xmath164}$ ] plus @xmath2 . the simplest implementation of @xmath186 uses an array of pairs @xmath187 sorted by @xmath169 in ascending order and by @xmath188 in the reverse order in the same time . this is possible based on lemma [ concave ] . the first block of instructions can be performed in @xmath189 by simply shifting all elements such that @xmath190 to the right , and modifying @xmath169 by iterating over all elements of @xmath186 . the next block ( computing of @xmath191 ) can be computed with the same complexity , as @xmath192 is simply the next element after @xmath169 in the ordered array . finally , every instruction in the last block can be performed in @xmath189 by iterating over all its elements . in this implementation @xmath186 is represented by global variables @xmath193 and a bst of values @xmath194 where @xmath188 is the key . an entry @xmath187 in this structure is represented as @xmath195 so obtaining an element of @xmath186 may take linear time . the advantages of this structure can be easily seen when analyzing the first block of the code . the division of @xmath186 by the value of @xmath188 ( at first @xmath69 ) can be achieved in @xmath196 . shifting all elements of those subsets can then be done in a constant time by modifying first elements of those sets . adding the extra node also requires @xmath196 operations . removing nodes with the first coordinate @xmath197 is obviously done in @xmath40 in total . identifying nodes with the first coordinate @xmath198 might seem problematic . it is , however , known that for the least @xmath188 the value of @xmath169 is equal to @xmath199 . relevant nodes can be , therefore , removed from the back in @xmath40 . adding @xmath91 to the second coordinate of each node is done by adding it to global variable @xmath193 . performance of the algorithm depends on the choice of @xmath186 data structure . theoretic bounds for computational complexity are , however , not sufficient to argue about performance of these two options . the first reason is that the each operation in @xmath185 algorithm is much faster than in @xmath1 in terms of number of instructions . secondly , hardware architectures provide solutions in which iterating over large tables is accelerated . finally , the algorithm does reach its theoretical bound only if many points concentrate on a small interval . a gap of size @xmath201 between two points completely cleans @xmath186 data structure . all that is shown in numerical tests . to measure performance we have used a single core of amd athlon ii x4 605e processor clocked at 2.3ghz with 8 gb of memory . the results are presented in figs.[fig1]-[fig2 ] . dirac deltas randomly distributed over @xmath202 $ ] . the plot shows how the time of computation depends on @xmath203 . for each input size 100 independent tests were executed to demonstrate how sensitive the algorithms are to input data distribution . results of @xmath204 algorithm are depicted as red dots , and results of @xmath6 algorithm as blue dots.,width=547 ] dirac deltas with a distributed over a large domain , i.e. distance between each two masses is larger than @xmath201 . in this case both algorithms are in fact linear , as the funcdescription structure has at most two elements . the plot demonstrates the overhead of using bst structures . results of @xmath204 algorithm are depicted as red dots , and results of @xmath6 algorithm as blue dots.,width=547 ] * acknowledgments * + jj was supported by the international ph.d . projects program of foundation for polish science operated within the innovative economy operational program 2009 - 2015 funded by european regional development fund ( ph.d . program : mathematical methods in natural sciences ) . am - c was supported by the erc starting grant `` biostruct '' no . 210680 and the emmy noether program of the german research council ( dfg ) . l. ambrosio and g. crippa . _ existence , uniqueness , stability and differentiability properties of the flow associated to weakly differentiable vector fields _ in transport equations and multi - d hyperbolic conservation laws , lect . notes unione mat . springer 5 : 357 , 2008 . , _ an introduction to the nonlinear boltzmann - vlasov equation _ , in kinetic theories and the boltzmann equation , springer , berlin , lecture notes in math . 1048 : 60110 , 1981 . , , 5 : 873900 , 2012 . . variational particle schemes for the porous medium equation and for the system of isentropic euler equations _ math . 44 : 133166 , 2010 . , springer - verlag , berlin , 2009	in this paper numerical methods of computing distances between two radon measures on @xmath0 are discussed . efficient algorithms for wasserstein - type metrics are provided . in particular , we propose a novel algorithm to compute the flat metric ( bounded lipschitz distance ) with a computational cost @xmath1 . the flat distance has recently proven to be adequate for the escalator boxcar train ( ebt ) method for solving transport equations with growth terms . therefore , finding efficient numerical algorithms to compute the flat distance between two measures is important for finding the residual error and validating empirical convergence of different methods . * keywords : * metric spaces , flat metric , wasserstein distance , radon measures , optimal transport , linear programming , minimum - cost flow
the primordial perturbation in the inflationary epoch is widely believed to be the primary origin of the formation of the large - scale structure in the universe @xcite . the generation of the curvature perturbation @xmath0 in the original inflation scenario is due to the inflaton perturbation @xmath1 that exits the horizon during inflation . the curvature perturbation @xmath0 in the original single - field inflation scenario is generated from the inflaton perturbation @xmath2 , which remains constant after the horizon exit . in the curvaton scenario , @xmath0 is instead generated after inflation from the isocurvature perturbation of a curvaton field @xcite ( for a recent review see ref . @xcite ) . in this paper we consider a hybrid potential for the curvaton scenario and study both the original ( oscillating ) @xcite and the inflating curvaton @xcite . the hybrid potential is very important in cosmological model building and a vast variety of models have been considered @xcite . this paper presents the first attempt in choosing the hybrid potential @xmath3 for the curvaton mechanism . in this paper we try to keep the discussion as simple as possible . we introduce three parameters @xmath4 , @xmath5 and @xmath6 to measure the model dependence of the scenario , and avoid highly model - dependent arguments . since the non - trivial evolution after the waterfall @xcite may ruin the simplicity of the model , we assume matter - like evolution ( @xmath7 ) that starts just after the phase transition . the phase transition is called `` trigger '' in this paper because the phase transition triggers the oscillation of the waterfall field @xmath8 . a crucial difference from the usual curvaton is that the source of the perturbation is not the perturbation of the oscillating ( waterfall ) field ( because @xmath9 for large scale ) but the perturbation of the slow - rolling field ( @xmath10 ) @xcite , which is not the inflaton ( @xmath11 ) of the primordial inflation but it is the field which triggers the phase transition . the characteristics of this scenario are shown in fig . [ fig : usual ] . ) does not happen if @xmath12 . as is discussed in the text , @xmath13 is determined by the model parameters . ] let us first consider the basic curvaton scenario . the initial perturbation of the curvaton @xmath8 is given by @xmath14 this corresponds to the perturbation @xmath15 in fig.[fig : usual ] . if the oscillation starts at @xmath16 , @xmath17 ( in the figure ) is exactly determined by the curvaton mass @xmath18 . then there is no perturbation which corresponds to @xmath19 . obviously , both @xmath20 and @xmath19 can source the curvaton perturbation , but in the basic scenario the source perturbation is almost entirely coming from @xmath15 . the total curvature perturbation during the curvaton oscillation is given by @xmath21 where we introduced the ratio @xmath22 . here @xmath23 is negligible . to calculate the curvaton perturbation , what we need is the quantity @xmath24 at the beginning of the oscillation because @xmath24 , defined by @xmath25 becomes constant during the ( quasi)sinusoidal oscillation . the new idea in this paper is that @xmath24 _ is dominated by the perturbation _ @xmath19 . in the basic ( non - hybrid ) scenario , the perturbation caused by @xmath26 does not dominate the total curvature perturbation . such modulated ( inhomogeneous ) interaction is not considered in this paper . ] the reason is as follows . the source of @xmath19 in the basic curvaton scenario is @xmath27 , and @xmath18 is not a constant when the potential is not quadratic . assuming that the sinusoidal oscillation starts at @xmath28 , the perturbation at the beginning of the oscillation is given by @xmath29 here we consider @xmath30 where , in contrast to the inflation phase , @xmath31 in the radiation dominated universe , which dominates the evolution of @xmath32 . assuming a polynomial function @xmath33 , we find @xmath34 and @xmath35 where the quadratic potential gives @xmath36 , which always leads to @xmath37 . for @xmath38 , we find @xmath39 the above perturbation is comparable to the perturbation sourced by @xmath15 in the basic curvaton scenario . therefore , this perturbation does not dominate the curvature perturbation , but may change its non - gaussianity . as the result , the deviation from the quadratic potential ( @xmath38 ) may cause @xmath40 , but this perturbation can not dominate over the total curvature perturbation . in the original curvaton scenario , the above dynamics has been included in the function @xmath41 introduced in ref . ( see also ref.@xcite ) to find a significant enhancement of the curvaton perturbation , consider a hybrid potential that is given by @xmath42 the potential during slow - roll is @xmath43 . the hybrid potential usually leads to generation of cosmic strings at the phase transition ( @xmath44 ) , but in this curvaton scenario the strings do not lead to a serious cosmological problem because the typical energy scale is much lower than the cosmological bound . suppose that the `` rolling field '' ( @xmath45 ) rolls from far away ( @xmath46 ) and the sinusoidal oscillation of the `` waterfall field '' ( @xmath8 ) starts suddenly after the phase transition . in this model , @xmath47 is determined by the `` rolling field '' reaching the waterfall at @xmath48 . , where is the critical value which triggers the phase transition . therefore , the perturbation @xmath19 @xcite is given by @xmath49 where @xmath50 denotes the velocity in the radiation - dominated universe evaluated at the waterfall . on the other hand , the initial value of the curvaton oscillation is given by @xmath51 , which means that the conventional source of the perturbation ( @xmath15 in fig . [ fig : usual ] ) vanishes in this model . for simplicity , assume that the energy density of the curvaton oscillation evolves as @xmath7 just after the phase transition . in this scenario , the perturbation of the `` rolling field '' ( @xmath10 ) is converted into @xmath19 to cause the perturbation @xmath52 . in this model , the transfer of the perturbation @xmath53 sources the curvaton perturbation @xmath54 . the modulation of the phase transition ( @xmath55 ) causes the perturbation of the curvaton density ; @xmath56 where @xmath57 is assumed for the oscillation . if @xmath45 rolls slowly before the phase transition , there is a significant enhancement of the curvaton perturbation . note that , in the above equation , @xmath50 is not the conventional slow - roll velocity which is usually given by @xmath58 . this field has a significant acceleration @xmath59 due to @xmath60 in the radiation - dominated universe . moreover , the typical expansion given by @xmath61 shows that the relation between @xmath62 and @xmath63 is obviously non - linear . this kind of non - linearity does not appear in the case of the standard quasi - de sitter inflation ( with @xmath45 the inflaton ) , since @xmath64 is usually assumed for @xmath45 to undergo slow - roll . however , @xmath64 is not true in the radiation - dominated background . for later convenience , we consider the approximation ; @xmath65 where @xmath66 denotes the hubble parameter at the phase transition . here @xmath67 is the reduced planck mass . the deviation from slow - roll is measured by @xmath68 . we assume @xmath69 to avoid extreme situations . may hold even if the field @xmath45 started oscillating near the waterfall . here the breakdown of slow - roll occurs first , then fast - roll may follow . there could be a breakdown of fast - roll if @xmath70 ( @xmath5 is defined in eq . ( [ beta ] ) ) , where @xmath45 starts oscillating but we assume the oscillation soon stops at the waterfall . assuming that fast - roll breaks down at @xmath71 , @xmath45 gains the kinetic energy @xmath72 . comparing @xmath73 with the ( possible ) slow - roll kinetic energy at the waterfall ; @xmath74 ^ 2 $ ] , we find approximately @xmath75 . note that in eq . ( [ eq_convert ] ) the conventional slow - roll is not assumed . ] in some extreme models the non - gaussianity could be affected , but in the present scenario the effect is at most @xmath76 . in the radiation - dominated universe and for the quadratic potential , @xmath77 has been derived in ref . the opposite ( @xmath78 ) corresponds to the inflating curvaton scenario . an intermediate scenario where @xmath79 is also discussed in this paper . considering the oscillation of @xmath8 as in the basic curvaton scenario , the density perturbation of the curvaton is @xmath80 from eq.([eq_convert ] ) , we find @xmath81 where @xmath82 denotes the slow - roll parameter of the rolling field @xmath45 . for later convenience , we introduce a new parameter @xmath5 that is defined by @xmath83 where @xmath84 is the possible soft - breaking mass that usually appears in supergravity when the f - term is given by @xmath85 . it is evident that @xmath86 is motivated by supergravity , and @xmath87 measures the deviation from the naive setup . @xmath88 is inevitable if the usual supersymmetry breaking sector ( which is not related to the curvaton potential ) gives the mass term with @xmath70 . in both cases @xmath5 is highly model - dependent . ] @xmath89 and @xmath86 are possible at the same time if the universe is dominated by the radiation @xcite . let us consider the spectrum @xmath90 for the slow - rolling field , where @xmath91 is the hubble parameter during the primordial inflation . defining the energy ratio at the beginning of the @xmath8 oscillation as @xmath92 we find at @xmath47 ; @xmath93 again , @xmath94 is possible when @xmath95 . the spectrum of the curvaton perturbation is given by @xmath96 here the evolution of @xmath63 is neglected because of the slow - roll assumption . defining the gravitational decay constant as @xmath97 , and considering the actual decay rate @xmath98 , where @xmath99 measures the deviation from the gravitational decay , we find that the hubble parameter at the decay is given by @xmath100 which leads to the density ratio @xmath101 where the evolution is possible until @xmath102 . here the value of @xmath103 ( at the beginning of the oscillation ) is determined by the initial condition . although the situation depends on the choice of the model , @xmath104 may cause significant interaction between @xmath45 and the radiation , which could alter the potential . in this sense , the potential used in the above argument is the effective potential approximated at the point very close to the waterfall and it includes all these corrections . the spectral index is @xmath105 and the non - gaussianity is @xmath106 for @xmath107 , as in the usual curvaton scenario . here @xmath108 and @xmath82 are defined at the horizon exit . as a second example , we consider the fast decay scenario . again , we find the spectrum of the perturbation given by @xmath109 here the fast - decay assumption leads to @xmath110 . to quantify the ratio , we introduce a parameter defined by @xmath111 where @xmath112 corresponds to the instant decay . we find @xmath113 as in the usual curvaton scenario , the phase transition at @xmath114 leads to enhanced non - gaussianity . the model predicts that , whenever @xmath115 is natural in the hybrid curvaton scenario , the instant decay may affect non - gaussianity . this is not mandatory in the slow - roll scenario , but could be mandatory if slow - roll breaks down before the waterfall . we are going to consider this possibility in the next section , although a more exact calculation requires numerical study . since the radiation energy density is decreasing while @xmath116 is constant , the natural idea is that @xmath117 triggers the significant variation of @xmath45 and eventually it causes the phase transition . because the contribution of @xmath45 to the scalar potential is small ( i.e , @xmath118 ) , the kinetic energy of @xmath45 is not significant even if the slow - roll is violated before the waterfall . to avoid an extreme situation , our modest assumption in this section is @xmath119 . on the other hand , the perturbation @xmath63 leaves horizon during the primordial inflation , when the slow - roll parameters of @xmath45 are supposed to be much smaller than unity . we thus find that the spectral index is small , @xmath120 . the problem in the slow - roll scenario is that , since the variation of @xmath45 ( i.e , @xmath121 ) is negligible , we always have to assume careful tuning of the initial condition @xmath122 . this tuning is relaxed if slow - roll breaks down just before the phase transition . moreover , since @xmath123 is justified in supergravity , there is the _ important prediction _ that @xmath124 in the hybrid curvaton model . to look into more details , consider the evolution in the radiation - dominated universe . this immediately leads to the equation of motion @xmath125 where the hubble parameter evolves as @xmath126 before fast - roll , we assume @xmath127 . substituting @xmath128 , we find for the radiation dominated universe @xcite @xmath129 which has an oscillating solution when @xmath130 according to ref . @xcite , there can be a cancellation that leads to @xmath131 during radiation domination . therefore , a conceivable set - up of the model is @xmath132 with fast - roll starting when @xmath133 . at this moment ( @xmath134 ) , we find @xmath135 to capture the evolution after the breaking of slow - roll , we introduce @xmath136 , which is determined by the evolution thereafter . @xmath137 means that the @xmath8-oscillation starts immediately after the breaking of slow - roll . note that a non - trivial evolution of the waterfall field can be described by @xmath138 , which we will not consider . there could be a small - scale perturbation of @xmath139 , but this is not the topic of this paper . for @xmath140 , the @xmath8-oscillation starts in the radiation - dominated universe . this means that @xmath141 is given by @xmath142 assuming fast decay ( @xmath143 ) and @xmath86 , the slow - roll breakdown just before the waterfall may produce naturally the coincidence @xmath144 . therefore , the scenario gives rise to @xmath144 , which can affect non - gaussianity . one may suspect that a large initial value ( @xmath145 ) can easily cause a long - time rolling that leads to the curvaton domination before the phase transition . this speculation is true . in this respect , the scenario more or less depends on the initial condition . @xmath146 with @xmath86 may lead to the opposite scenario , in which domination by the curvaton starts _ before _ the phase transition . this scenario of the inflating curvaton @xcite will be discussed in the next section . in the above scenario with @xmath86 , long - time oscillation of the waterfall field is not needed for the curvaton domination . the waterfall field @xmath8 can decay after a few oscillations , _ in contrast to the standard curvaton scenario _ , in which a small decay rate is needed for the curvaton domination . the spectrum of the curvature perturbation is obtained in the same way as in the previous ( slow - roll ) scenario ; @xmath147 where the ratio @xmath148 behaves like a constant when @xmath149 is quadratic . otherwise , the result is highly model - dependent . note that the deviation from @xmath150 might be significant at the waterfall if the evolution is already far from slow - roll ( i.e , when @xmath151 ) . in this case , the non - linear evolution of @xmath45 and @xmath63 is significant and it requires further study . here , considering a model with the modest evolution @xmath150 and @xmath86 ( as motivated by supergravity ) , we expect @xmath152 . it is possible to realize the inflating curvaton scenario if the oscillation starts after curvaton domination . this scenario always gives @xmath153 by definition , and helps to explain the spectral index and its running by reducing the number of e - foldings required for the primordial inflation @xcite . the number of e - foldings during the curvaton inflation ( @xmath154 ) is determined by the initial condition @xmath155 and @xmath5 ; @xmath156 where @xmath157 denotes the beginning of the curvaton inflation . in this section we assume @xmath76 couplings for the hybrid potential ( @xmath158 ) just for simplicity . considering the bound @xmath159 and @xmath160tev ( i.e , @xmath161tev ) , we find from eq . ( [ n2 ] ) that the maximum number of @xmath154 is given by @xmath162 slow - roll ( @xmath163 ) requires @xmath164 if we demand @xmath165 . we are not excluding this possibility , but it would be interesting to relax this tuning using a fast - roll potential ( @xmath166 ) . in this case we find @xmath167 since the ratio @xmath148 does not evolve in the case of a quadratic potential , we obtain @xmath168 where the relation @xmath169 was used . here @xmath170 denotes the observable spectrum from the cmb . if we consider the stochastic initial condition @xcite , we find @xmath171 . here @xmath172 is used because @xmath76 couplings are assumed . therefore , the stochastic condition leads to @xmath173 this result is in sharp contrast to the pngb inflating curvaton @xcite , in which the stochastic condition puts an _ upper _ bound for @xmath154 . an interesting application of this scenario is that @xmath174 can be related to the dynamical scale of supersymmetry breaking while @xmath86 is due to the supergravity action . considering @xmath175gev , we find ( the upper bound is different from eq.([35beta ] ) because the scale of @xmath174 is different . ) @xmath176 and for @xmath177gev we find @xmath178 as in the standard curvaton scenario , the spectral index is given by @xmath179 where @xmath180 ( for the primordial inflaton ) and @xmath32 ( for the slow - roll curvaton field ) are calculated at the horizon exit . if the primordial inflation is of the chaotic type with potential @xmath181 , then @xmath182 . according to observations , @xmath183 @xcite , which leads to @xmath184 for negligible @xmath32 . in the oscillating curvaton scenario ( @xmath185 ) the spectral index excludes @xmath186 , and gives for the running of the spectral index @xmath187 the situation is quite different in the inflating curvaton scenario . because of the second inflation @xmath188 , @xmath186 is not ruled out , while the running of the spectral index is enhanced when @xmath189 , which can have observational impact @xcite . one of the advantages of the curvaton idea is that the sector , which gives rise to primordial inflation ( inflaton sector ) , can be decoupled from the sector which gives rise to the curvature perturbation ( curvaton sector ) . this can naturally accommodate the stability of the hierarchy in scales , which is a feature of the curvaton scenario . moreover , in the curvaton scenario , the inflaton sector does not even need to feature a fundamental scalar field at all gravity in the inflationary cosmology was pioneered by starobinsky @xcite . ] in that sense , the curvaton scenario liberates the inflaton sector from many serious constraints @xcite . although the inflaton sector has been investigated in the context of various theoretical models , including the hybrid - type potential for the inflaton field , the curvaton sector is usually considered merely with the simple quadratic potential . this gives somewhat `` biased '' information about the possibilities for inflationary cosmology . in that sense , our paper is an effort to expand the state of the art . furthermore , the hybrid potential , which is the target in this paper , is well motivated in particle physics without being necessarily tied to the inflaton . for example , the hybrid structure may appear in guts or brane - brane interactions @xcite while it can well explain the differences in scales @xcite . finally , in the hybrid curvaton , the source mechanism of the curvature perturbation is utterly different from the original curvaton scenario . in that sense the basic formulation is brand - new . to be more specific , we considered in this paper a new possibility of the curvaton scenario ; we studied oscillating , inflating and non - slow - roll scenarios for the hybrid curvaton . for the slow - roll scenarios we found significant enhancement of the curvature perturbation , while in the non - slow - roll scenarios we found that the tuning requirements for the initial condition are relaxed . the inflating curvaton predicts distinguishable running of the spectral index , which can not be attained in the usual inflation scenario except for a few extended models . there is a variety of models possible for the hybrid curvaton . for example , one can consider the curvaton analogues of smooth hybrid inflation @xcite or shifted hybrid inflation @xcite . since the purpose of this paper is not the model building but the basic formulation of the brand - new idea , we avoided highly model - dependent arguments and focused on the simplest model that is useful to explain the essential idea of the scenario . we have considered @xmath190 because it is the natural choice . however , the scenario with @xmath191 is interesting as well . a large @xmath4 would reduce the spectrum of the curvature perturbation , but the non - linear evolution might enhance non - gaussianity , in a way that has not been considered yet . however , the calculation requires numerical study . we have also considered @xmath86 ( @xmath5 is defined in eq . ( [ beta ] ) ) because this is well motivated by supergravity . @xmath5 is generically a model - dependent parameter and , were it not for the supergravity motivation , there would be no reason to believe that @xmath86 . finally , we considered that the oscillation starts immediately after the end of the slow - roll evolution to avoid the complexity related to the dynamics of the waterfall . however , in some cases this dynamics could be non - trivial , which may require numerical calculations . indeed , the evolution of the waterfall is not easily understood by analytical calculation only @xcite even without a radiation background . we thank david h. lyth for valuable discussions . kd is supported ( in part ) by the lancaster - manchester - sheffield consortium for fundamental physics under stfc grant st / j000418/1 . kk is partly supported by the grant - in - aid for the ministry of education , culture , sports , science and technology , government of japan nos . 21111006 , 22244030 , 23540327 , and by the center for the promotion of integrated science ( cpis ) of sokendai ( 1hb5806020 ) . 1 d. h. lyth and a. r. liddle , _ the primordial density perturbation _ , cambridge university press , 2009 . s. mollerach , phys . d * 42 * ( 1990 ) 313 ; a. d. linde and v. f. mukhanov , phys . d * 56 * , 535 ( 1997 ) [ astro - ph/9610219 ] ; d. h. lyth and d. wands , phys . b * 524 * , 5 ( 2002 ) [ hep - ph/0110002 ] ; t. moroi and t. takahashi , phys . b * 522 * , 215 ( 2001 ) [ erratum - ibid . b * 539 * , 303 ( 2002 ) ] [ hep - ph/0110096 ] ; k. enqvist and m. s. sloth , nucl . b * 626 * ( 2002 ) 395 [ arxiv : hep - ph/0109214 ] . a. mazumdar and j. rocher , phys . * 497 * ( 2011 ) 85 [ arxiv:1001.0993 [ hep - ph ] ] . k. dimopoulos , k. kohri , d. h. lyth and t. matsuda , arxiv:1110.2951 [ astro-ph.co ] . k. furuuchi and c. -m . lin , arxiv:1111.6411 [ hep - ph ] . d. h. lyth , arxiv:1107.1681 [ astro-ph.co ] ; d. h. lyth , jcap * 1107 * , 035 ( 2011 ) [ arxiv:1012.4617 [ astro-ph.co ] ] . k. dimopoulos and m. axenides , jcap * 0506 * , 008 ( 2005 ) [ hep - ph/0310194 ] . t. matsuda , phys . d * 68 * , 047702 ( 2003 ) [ hep - ph/0302253 ] . t. matsuda , class . grav . * 26 * , 145011 ( 2009 ) . [ arxiv:0902.4283 [ hep - ph ] ] ; t. matsuda , jcap * 0812 * , 001 ( 2008 ) . [ arxiv:0811.1318 [ hep - ph ] ] . t. matsuda , jhep * 0707 * , 035 ( 2007 ) . [ arxiv:0707.0543 [ hep - th ] ] . d. h. lyth , phys . b * 579 * ( 2004 ) 239 [ hep - th/0308110 ] . d. h. lyth , c. ungarelli and d. wands , phys . d * 67 * , 023503 ( 2003 ) [ astro - ph/0208055 ] ; k. dimopoulos , g. lazarides , d. lyth and r. ruiz de austri , phys . d * 68 * , 123515 ( 2003 ) [ hep - ph/0308015 ] . m. kawasaki , t. kobayashi , f. takahashi , [ arxiv:1107.6011 [ astro-ph.co ] ] . k. a. malik , d. wands , class . * 21 * , l65-l72 ( 2004 ) . [ astro - ph/0307055 ] . p. binetruy , g. r. dvali , phys . lett . * b388 * , 241 - 246 ( 1996 ) . [ hep - ph/9606342 ] ; e. halyo , phys . * b387 * , 43 - 47 ( 1996 ) . [ hep - ph/9606423 ] . t. matsuda , jcap * 0311 * , 003 ( 2003 ) . [ hep - ph/0302078 ] ; t. matsuda , phys . * b423 * , 35 - 39 ( 1998 ) . [ hep - ph/9705448 ] . d. h. lyth and t. moroi , jhep * 0405 * , 004 ( 2004 ) [ hep - ph/0402174 ] . e. komatsu _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 192 * , 18 ( 2011 ) . e. d. stewart , phys . b * 391 * , 34 ( 1997 ) [ hep - ph/9606241 ] ; e. d. stewart , phys . d * 56 * , 2019 ( 1997 ) [ hep - ph/9703232 ] ; s. m. leach , i. j. grivell and a. r. liddle , phys . d * 62 * , 043516 ( 2000 ) [ astro - ph/0004296 ] . k. kohri , c. -m . lin and d. h. lyth , jcap * 0712 * , 004 ( 2007 ) [ arxiv:0707.3826 [ hep - ph ] ] . g. lazarides and c. panagiotakopoulos , phys . d * 52 * ( 1995 ) r559 [ arxiv : hep - ph/9506325 ] . r. jeannerot , s. khalil , g. lazarides and q. shafi , jhep * 0010 * ( 2000 ) 012 [ arxiv : hep - ph/0002151 ] ; r. jeannerot , s. khalil and g. lazarides , jhep * 0207 * ( 2002 ) 069 [ arxiv : hep - ph/0207244 ] . a. a. starobinsky , phys . b * 91 * ( 1980 ) 99 . c. herdeiro , s. hirano and r. kallosh , jhep * 0112 * , 027 ( 2001 ) [ hep - th/0110271 ] . k. dimopoulos and d. h. lyth , phys . d * 69 * ( 2004 ) 123509 [ arxiv : hep - ph/0209180 ] . j. garcia - bellido , hep - ph/0002256 .	we study both oscillating and inflating curvaton scenarios when the curvaton mechanism is caused by a hybrid potential . the source of the curvature perturbation is the inhomogeneous phase transition that causes the modulation of the onset of the oscillation . for the supergravity - motivated curvaton there is a possibility of finding natural coincidence of the energy density that is needed to affect non - gaussianity in the curvaton scenario .
the extent to which the non - classical properties of one - mode field states survive in the presence of noise and losses was investigated since the early years of quantum optics @xcite . a routine operation like the transmission of light beams through an optical fiber could produce a substantial degradation of their non - classical properties . as an example , it was found that squeezing properties are altered by admixture with thermal noise and disappear completely for values of thermal mean photon occupancy exceeding the threshold @xmath0 @xcite . from a more recent quantum - information perspective , a lot of work is concentrated on correlations such as entanglement and discord in multi - partite systems @xcite . while correlations associated with entanglement @xcite are defined in connection to global transformations of bipartite quantum states , the concept of quantum discord arises from local ( marginal ) actions and measurements performed on one subsystem @xcite . its definition contains an optimization over the set of all one - party measurements , which in the case of mixed states could be a challenging problem . note that in the pure - state case , entanglement and discord coincide and therefore they measure the total amount of correlations . in the mixed - state case , quantum discord is a measure of quantumness whose relation to entanglement is not a simple one . a survey of recent progress and applications of classical and quantum correlations quantified by quantum discord and other measures can be found in refs . @xcite . when states of a two - party quantum system need to interact with a noisy channel , a drastic modification of their quantum correlations is expected to occur @xcite . for instance , it was found that quantum and classical correlations for a system of two qubits evolving in markovian dephasing channels can display different dynamics @xcite . quite recently , the effect of local noisy channels on quantum correlations in finite - dimensional quantum systems was investigated @xcite . it was found that while entanglement does not increase under local channels , other correlations can become larger when the input state is not pure . in the continuous - variable settings , a similar behaviour was recently noticed for the gaussian discord of two - mode mixed states under single - mode gaussian dissipative channels @xcite . local gaussian thermal and phase - sensitive reservoirs modify the entanglement properties of two - mode gaussian states , as interestingly is pointed out in refs . evolutions of more general gaussian correlations were also investigated in refs . @xcite . in this work we analyze the decay of quantum correlations of the field in a two - mode gaussian state due to the interaction of the modes with separate thermal baths . we focus on two measures of quantum correlations , namely , the entanglement of formation ( ef ) and the quantum discord , and evaluate them for a damped two - mode squeezed thermal state ( sts ) @xcite . our choice of this important particular class of gaussian states is motivated by the recent result @xcite that for an sts , the exact discord according to its original definition @xcite is achieved with an optimal measurement which is gaussian . consequently , what was called the gaussian discord and was derived in refs . @xcite is actually the _ exact _ discord . on the other hand , in the interesting paper @xcite it is shown that for an sts , a gaussian character of the discord implies that the ef is also gaussian . accordingly , the gaussian ef written explicitly in our paper @xcite is equally an exact result . as such , when dealing with a dissipative evolution that preserves the stss , we have the rare privilege to fully describe the decay of two types of correlations by analytic means . our paper is structured as follows . in section 2 we recapitulate several properties of an sts : the covariance matrix , the simon separability criterion @xcite , the entanglement of formation @xcite as a measure of inseparability , and the quantum discord as derived in refs . @xcite . in section 3 the system of interest ( field @xmath1 environment ) is specified in order to write and solve the quantum optical master equation . special attention is then paid to the solution for an input sts . section 4 deals with a damped sts for modes coupled with two identical local thermal baths . we here derive the time at which the entanglement sudden death occurs and study the evolution of the discord . section 5 is dedicated to another interesting configuration of the system : only one mode is in contact with a thermal bath . we find that all the pure gaussian states lose their entanglement at the same time which depends on the reservoir only . in the case of a mixed input state , the discord defined by local measurements on the attenuated mode is increased above its initial value . our conclusions are drawn in sec . in this section we review shortly the above - mentioned examples of gaussian correlations , with emphasis on those of an sts . we consider two - mode gaussian states and let us denote the photon annihilation operators of the modes by @xmath2 and @xmath3 as shown in refs . @xcite , an sts is the result of the action of a two - mode squeeze operator , @xmath4 } , \quad \left ( r>0,\;\ ; \phi\in ( -\pi,\pi ] \right),\end{aligned}\ ] ] on a two - mode thermal state with the mean photon occupancies @xmath5 and @xmath6 : @xmath7 its covariance matrix ( cm ) has the following block structure @xcite : @xmath8 in eq . ( [ cv ] ) , @xmath9 denotes the 2@xmath102 identity matrix and @xmath11 is the 2@xmath102 symmetric matrix @xmath12 recall that the cm of an sts , eqs . ( [ cv ] ) and ( [ mc ] ) , has the standard - form parameters @xcite : @xmath13 ^ 2 + \left ( \bar{n}_2+\frac{1}{2}\right ) [ \sinh(r)]^2 , \nonumber \\ b_2&=&\left ( \bar{n}_1+\frac{1}{2}\right)[\sinh(r)]^2+\left ( \bar{n}_2 + \frac{1}{2}\right ) [ \cosh(r)]^2 , \nonumber \\ c&=&(\bar{n}_1+\bar{n}_2 + 1)\sinh(r ) \cosh(r ) . \label{par}\end{aligned}\ ] ] in many applications one can take advantage of a formal definition of an sts , as being an undisplaced and unscaled two - mode gaussian state described by three standard - form parameters : @xmath14 . if @xmath15 , then these parameters must fulfill the uncertainty inequality @xmath16 if @xmath17 , then one has to interchange the parameters @xmath18 and @xmath19 in eq . ( [ state ] ) @xcite . the standard form of its cm is given by eq . ( [ cv ] ) with the @xmath20 matrix @xmath11 written for @xmath21 , i.e. , becoming proportional to the pauli matrix @xmath22 . within this formal treatment , eq . ( [ sts ] ) and its companions , eqs . ( [ cv]) ( [ par ] ) , represent a parametrization of an sts with a clear experimental relevance . it is known that , according to williamson s theorem @xcite , the cm of a two - mode gaussian state can be diagonalized by a symplectic transformation . we get thus an important ingredient in describing the state , namely , the symplectic eigenvalues of the cm . for an sts they are @xcite : @xmath23 . \label{se}\end{aligned}\ ] ] in the parametrization ( [ par ] ) , we get @xmath24 . it is worth mentioning simon s separability criterion for two - mode gaussian states @xcite . it was proven that preservation of the non - negativity of the density matrix under partial transposition is not only a necessary @xcite , but also a sufficient condition for the separability of two - mode gaussian states @xcite . accordingly , a two - mode gaussian state is separable when the condition @xmath25 is met . we have denoted by @xmath26 the symplectic eigenvalues of the cm corresponding to the partial transpose of the density matrix . for an sts one finds : @xmath27 . \label{pse}\end{aligned}\ ] ] a simplified form of the separability condition for an sts that we shall use in what follows reads @xcite : @xmath28 before proceeding , let us note that , apart from the vacuum state , the only undisplaced and unscaled pure two - mode gaussian states are the squeezed vacuum ones , @xmath29 , and they belong to the set of stss . the pure - state case is characterized by the identities @xmath30 and @xmath31 . equations ( [ se ] ) and ( [ pse ] ) give now @xmath32 . in the following , we shall recall two measures of quantum correlations for an sts , namely , the ef and the quantum discord . the ef is defined as an optimization over all the pure - state decompositions of the given state @xcite : @xmath33 . \label{od}\end{aligned}\ ] ] in the expression above , we have denoted by @xmath34 the entanglement of the pure bipartite state @xmath35 . we here focus on the case of an sts @xmath36 and recall that an expression for its ef could be obtained when restricting the optimization in eq . ( [ od ] ) to gaussian pure - state decompositions only . for further convenience , we introduce the entropic function @xmath37 it was proven that the gaussian ef can be expressed in terms of the function @xmath38 as : @xmath39 in eq . ( [ expr - ef ] ) , the parameter @xmath40 is given in terms of the entries of the cm @xcite : @xmath41 here @xmath42 is the main symplectic invariant . for any squeezed vacuum state , @xmath43 so that @xmath44 , and thus its ef is equal to the von neumann entropy of the reduced one - mode thermal state , i. e. , @xmath45 . the difference between two classically equivalent definitions of the mutual information provides another measure of the total amount of quantum correlations in a quantum state , called discord @xcite . let us consider a bipartite state @xmath46 and write down its quantum mutual information , @xmath47 with @xmath48 being the von neumann entropy of the state @xmath49 . another quantum analogue of the mutual information is more complicated and depends on the influence on the first subsystem @xmath50 of the measurements made on the second subsystem @xmath51 . let us denote by @xmath52 a quantum measurement performed on the system @xmath51 . the final state of the subsystem @xmath50 after such a measurement on the subsystem @xmath51 leading to the outcome @xmath53 is @xmath54 in eq . ( [ meas - state ] ) , @xmath55 is the probability of the outcome @xmath53 : @xmath56 . the quantum conditional entropy , given the non - selective measurement @xmath57 , is a convex sum of von neumann entropies of the post - measurement states ( [ meas - state ] ) which is taken over all the possible outcomes : @xmath58 the quantum information gained about the subsystem @xmath50 by taking into account the minimal disturbance produced on it by any of all the possible measurements performed on the subsystem @xmath51 is the difference @xcite @xmath59 the quantum @xmath50-discord is then defined as follows @xcite : @xmath60 similarly , the quantum @xmath51-discord , which considers the local quantum measurements performed on the first subsystem is @xmath61 quite recently , the above - defined discord @xcite has been calculated for two - mode gaussian states under the approach of limiting the set of all one - party quantum measurements to the gaussian ones @xcite . we were thus provided with an analytic formula for what is called the gaussian discord . moreover , according to ref . @xcite , at least for the states analyzed here , namely , the stss , the gaussian discord is the _ exact _ discord . thus the quantum discords ( [ d1 ] ) and ( [ d2 ] ) turn out to have very simple expressions in terms of one - mode von neumann entropies : @xmath62 here @xmath63 is the entropic function ( [ h ] ) and the symplectic eigenvalues @xmath64 are given in eq . ( [ se ] ) . in addition , we have used the notations : @xmath65 note that , for symmetric stss ( @xmath30 ) , the identity @xmath66 holds and therefore @xmath67 . moreover , for pure two - mode gaussian states , we get @xmath68 and @xmath69 , i. e. , the discord and the entanglement coincide , as expected @xcite . we consider an arbitrary two - mode field state having the annihilation operators @xmath70 and the density operator @xmath49 . each mode is in contact with a local thermal bath . we denote the mean photon occupancies of the two thermal reservoirs by @xmath71 , respectively , and the corresponding damping rates by @xmath72 . in the interaction picture , the quantum optical master equation which describes this type of coupling is @xmath73 as in our recent work @xcite for the one - mode case , instead of the master equation ( [ me ] ) , we employ the equivalent differential equation for the two - mode characteristic function @xmath74\hat \rho(t)\}$ ] . here @xmath75 and @xmath76 are the weyl displacement operators of the modes : @xmath77 we finally find the solution : @xmath78 \nonumber\\ & & \times \exp \left [ -\left(\bar{n}_{r2}+\frac{1}{2}\right)\left ( 1-e^{-\gamma_2t}\right ) \|\lambda_2\|^2\right ] . \label{cf}\end{aligned}\ ] ] let us inspect the asymptotic behaviour of the solution ( [ cf ] ) of the master equation ( [ me ] ) . when we take @xmath79 in eq . ( [ cf ] ) , we get the characteristic function of the two - mode thermal state imposed by the two reservoirs : @xmath80 . \label{asympcf}\ ] ] note that this two - mode steady state , which is independent of the input state , is a product state without any correlations between the modes . given the structure of the time - dependent characteristic function ( [ cf ] ) , any input gaussian state preserves its gaussian form at any time during the mode damping . in particular , an initial sts remains an sts at any subsequent time . its evolving cm has the following standard - form entries : @xmath81 . \label{spt}\end{aligned}\ ] ] in view of eqs . ( [ spt ] ) , the cm of the damped sts becomes asymptotically diagonal : @xmath82 this means that the two - mode steady state is a product one , whose factors are precisely the single - mode thermal states conditioned by the corresponding reservoirs . thus we recover the previous general conclusion in the special case of an an initial sts . to sum up , any measure of correlations in the gaussian approach available for an sts , such as the entanglement of formation @xcite or the quantum discord @xcite , can readily be applied for a decaying sts on account of eqs . ( [ spt ] ) . what are we expecting to occur when a two - mode quantum state is subjected to a dissipative interaction as described by the master equation ( [ me ] ) ? in general , a substantial reduction of the non - classical properties of the state which entails a decrease of its quantum correlations such as entanglement and discord . more specifically , in the important particular case of two - mode gaussian states , we can notice from the very beginning an important difference between the ways in which these two measures of quantum correlations actually decay . indeed , on the one hand , according to condition ( [ sc ] ) , the entanglement of the input state is expected to vanish at a finite time . this process has been called the _ entanglement sudden death _ in the case of qubits @xcite . on the other hand , it is known that the only zero - discord two - mode gaussian states are the product ones @xcite . taking account of the time - dependent two - mode characteristic function ( [ cf ] ) , as well as of its steady - state form ( [ asympcf ] ) , we infer that only the latter describes a product state without any correlations between the modes . therefore , it is reasonable to believe that only asymptotically a damped two - mode gaussian state could lose all its correlations , both quantum and classical , measured by the gaussian discord . for the sake of simplicity and in order to get versatile analytic results , we consider here the particular case when the two local reservoirs are identical : @xmath83 and @xmath84 . in this case , the cm of an arbitrary damped two - mode gaussian state reads : @xmath85 here @xmath86 is the input cm and @xmath87 denotes the @xmath88 identity matrix . equation ( [ cm ] ) tells us that an input state with no local squeezing does not change its character during damping : for instance , a symmetric state remains symmetric and an sts evolves as a damped sts . we restrict ourselves now to this latter case . when employing the entries of the time - dependent cm ( [ cm ] ) in the separability condition ( [ sc ] ) , one finds a simple expression of the time required by a damped sts to reach the separability threshold : @xmath89 here @xmath90 is the smallest symplectic eigenvalue of the cm of the partially transposed input density matrix , which is given by eq . ( [ pse ] ) . we see that in the special case of zero - temperature baths , the entanglement disappears only asymptotically . in all other cases , the quantum - classical transition occurs at finite times , i. e. , it happens a _ sudden death of entanglement _ @xcite . ( black line ) and @xmath91 ( dashed red line ) for an input sts in interaction with two local identical thermal reservoirs . we have employed the following parameters . ( a ) the sts is characterized by the parameters @xmath92 , @xmath93 , @xmath94 and the reservoir by @xmath95 . ( b ) for the same input state we use @xmath96 . ( c ) we plot the ef and the discord @xmath67 ( dashed red line ) for an input pure state having the squeeze parameter @xmath94 . the reservoir is noisy with @xmath95 . , title="fig:",width=188 ] ( black line ) and @xmath91 ( dashed red line ) for an input sts in interaction with two local identical thermal reservoirs . we have employed the following parameters . ( a ) the sts is characterized by the parameters @xmath92 , @xmath93 , @xmath94 and the reservoir by @xmath95 . ( b ) for the same input state we use @xmath96 . ( c ) we plot the ef and the discord @xmath67 ( dashed red line ) for an input pure state having the squeeze parameter @xmath94 . the reservoir is noisy with @xmath95 . , title="fig:",width=188 ] ( black line ) and @xmath91 ( dashed red line ) for an input sts in interaction with two local identical thermal reservoirs . we have employed the following parameters . ( a ) the sts is characterized by the parameters @xmath92 , @xmath93 , @xmath94 and the reservoir by @xmath95 . ( b ) for the same input state we use @xmath96 . ( c ) we plot the ef and the discord @xmath67 ( dashed red line ) for an input pure state having the squeeze parameter @xmath94 . the reservoir is noisy with @xmath95 . , title="fig:",width=188 ] in figs . [ fig1](a ) and [ fig1](b ) we plot the evolution of the entanglement of formation , eq . ( [ expr - ef ] ) , as well as that of the quantum discords @xmath97 and @xmath91 , eqs . ( [ d1 ] ) and ( [ d2 ] ) , for an asymmetric mixed sts . the aspect of the plots follows closely our above remarks on the robustness of discord against noise in comparison with the fragility of entanglement . note that the discords @xmath97 and @xmath91 are very close and can be distinguished only for very different values of the thermal mean photon occupancies @xmath98 and @xmath99 . when the reservoirs are noisy , ( @xmath100 ) , both the ef and the discords @xmath101 are strongly diminished . contact with zero - temperature baths , as in fig . [ fig1](b ) , produces a slower decay of all correlations , which in this case disappear only asymptotically . in fig . [ fig1](c ) we consider an input pure state , namely , a two - mode squeezed vacuum state in contact with a noisy bath . at the time @xmath102 the entanglement and discord coincide , but their time developments look very different . notice that , in view of eq . ( [ cm ] ) , a thermalized two - mode squeezed vacuum state evolves into a symmetric sts having @xmath67 . for finite - dimensional quantum systems it was recently found that while entanglement does not increase under local channels , other correlations such as discord can become larger when the input state is not pure @xcite . in continuous - variable settings , a similar behaviour was noticed for gaussian discord of mixed two - mode states under one - mode gaussian dissipative channels @xcite . to investigate here such an interaction with analytic means and results , we consider an input sts having only the mode 1 in contact with a thermal reservoir . we specialize the master equation ( [ me ] ) to the values @xmath103 so that the standard - form entries ( [ spt ] ) of the damped cm become : @xmath104 by insertion of the time - dependent parameters ( [ time - dep ] ) into the separability condition ( [ sc ] ) one finds the time at which the ef of a damped sts vanishes : @xmath105 , \quad \left ( b_1-\frac{1}{2}\right ) \left ( b_2-\frac{1}{2}\right)-c^2 < 0 . \label{ts1}\ ] ] we specialize eq . ( [ ts1 ] ) to the case of a pure gaussian input @xmath106 the time of the death of entanglement ( [ ts1 ] ) is then independent of the input two - mode squeezed vacuum state , being determined only by the field - reservoir coupling : @xmath107 we have also checked that the time of the entanglement death has the same expression ( [ tc ] ) for an input squeezed vaccuum state with additional local squeezings on both modes . moreover , in our recent paper @xcite , we found that , for some classes of one - mode states displaying initially certain negativities of their glauber - sudarshan @xmath108 representation , @xmath109 is the ultimate time at which the @xmath108 function becomes positive due to the field interaction with a thermal reservoir . at the time ( [ tc ] ) it therefore occurs a sudden quantum - classical transition for some types of one - mode states , as well as for any two - mode squeezed vacuum state . as regards the evolution of the gaussian discord , we expect it to decay eventually very slowly and to vanish only asymptotically . indeed , according to eqs . ( [ time - dep ] ) , the cm of the damped sts has an asymptotically diagonal form : @xmath110 the steady state of the field is therefore the product of two single - mode thermal states : the state of the damped mode 1 , which is imposed by the thermal reservoir owing to their interaction , and that of the freely - evolving mode 2 , which is its reduced state remaining constant in time and thus equal to its input at @xmath102 . the case of a pure - state input deserves additional remarks . according to eqs . ( [ time - dep ] ) , although at the moment @xmath102 the three measures of quantum correlations ef , @xmath97 and @xmath91 coincide , they behave subsequently quite differently because an input two - mode squeezed vacuum state evolves into an asymmetric sts . figure [ fig2](c ) displays the evolution of the ef , as well as those of both discords @xmath97 and @xmath91 , which are all monotonic , as predicted in ref . however , the discord @xmath91 corresponding to local measurements performed on the damped mode 1 survives much longer than both the ef and the discord @xmath97 . the case of an initial mixed state can be tackled by using eqs . ( [ expr - ef ] ) , ( [ discord ] ) , and ( [ time - dep ] ) for obtaining the expressions of the ef and the discords @xmath97 and @xmath91 . we plot in figs . [ fig2](a ) and [ fig2](b ) their time evolution for the same input state , but with a noisy bath ( a ) and a zero - temperature reservoir ( b ) . an enhancement of @xmath91 is noticed in both panels ( a ) and ( b ) . the discord @xmath91 presents a clear maximum in the latter situation and is much enhanced with respect to its value at the moment @xmath102 . this can be interpreted as a creation of quantum correlations similar to those first explored for finite - dimensional systems @xcite . moreover , in the recent ref . @xcite it was found that an enhancement of the discord @xmath91 can be noticed even when the input gaussian state is separable . ( black line ) and @xmath91 ( dashed red line ) in a thermal bath acting on mode 1 . the input state is characterized by the squeeze parameter @xmath94 . the input thermal mean photon occupancies are @xmath92 , @xmath111 ( left and central panels ) and the reservoir has ( a ) @xmath95 and ( b ) @xmath96 . the state considered in the panel ( c ) is pure , while the reservoir has @xmath95.,title="fig:",width=188 ] ( black line ) and @xmath91 ( dashed red line ) in a thermal bath acting on mode 1 . the input state is characterized by the squeeze parameter @xmath94 . the input thermal mean photon occupancies are @xmath92 , @xmath111 ( left and central panels ) and the reservoir has ( a ) @xmath95 and ( b ) @xmath96 . the state considered in the panel ( c ) is pure , while the reservoir has @xmath95.,title="fig:",width=188 ] ( black line ) and @xmath91 ( dashed red line ) in a thermal bath acting on mode 1 . the input state is characterized by the squeeze parameter @xmath94 . the input thermal mean photon occupancies are @xmath92 , @xmath111 ( left and central panels ) and the reservoir has ( a ) @xmath95 and ( b ) @xmath96 . the state considered in the panel ( c ) is pure , while the reservoir has @xmath95.,title="fig:",width=188 ] in order to draw some conclusions on the effects produced by local dissipation on quantum correlations of an sts , we compare now the decay of entanglement and discord for the two situations studied above . figure [ fig3 ] displays our results for the ef ( blue curves ) and @xmath91 ( black plots ) in the cases of both one and two local identical reservoirs for a mixed sts ( panels ( a ) and ( b ) ) and for a pure gaussian state ( c ) . ( black curves ) for one local bath ( dashed lines ) and two local identical baths ( full lines ) for an input state with the squeeze parameter @xmath94 . the other parameters are : ( a ) @xmath92 , @xmath111 , @xmath96 ; ( b ) @xmath92 , @xmath111 , @xmath95 ; ( c ) an input pure state and a noisy bath with @xmath95.,title="fig:",width=188 ] ( black curves ) for one local bath ( dashed lines ) and two local identical baths ( full lines ) for an input state with the squeeze parameter @xmath94 . the other parameters are : ( a ) @xmath92 , @xmath111 , @xmath96 ; ( b ) @xmath92 , @xmath111 , @xmath95 ; ( c ) an input pure state and a noisy bath with @xmath95.,title="fig:",width=188 ] ( black curves ) for one local bath ( dashed lines ) and two local identical baths ( full lines ) for an input state with the squeeze parameter @xmath94 . the other parameters are : ( a ) @xmath92 , @xmath111 , @xmath96 ; ( b ) @xmath92 , @xmath111 , @xmath95 ; ( c ) an input pure state and a noisy bath with @xmath95.,title="fig:",width=188 ] we represent here the case of zero - temperature reservoirs ( panels ( a ) and ( c ) ) to show a better preservation of all correlations in comparison with the noisy bath considered in panel ( b ) . we can see that in all cases both gaussian discords @xmath97 and @xmath91 survive longer than the ef . this is expected because only asymptotically the damped gaussian state becomes a product one . thus the gaussian discord , which measures the whole amount of quantum and classical correlations , proves to be quite robust against dissipation in all the above - mentioned situations . however , only for a configuration with one local thermal bath , there is an enhancement of the discord @xmath91 . since this can be larger than the discord of the input state , it means that the field - reservoir interaction generates quantum and clasical correlations of the discord type . a final conclusion arising from fig . [ fig3 ] is quite interesting . in all the analyzed situations ( mixed or pure input states , noisy or zero - temperature reservoirs ) , the configuration with one local thermal bath performs better than that with two local identical baths . this is valid when analyzing both the magnitude of correlations and their preservation in time . 99 vourdas a and weiner r m 1987 _ phys . rev . a _ * 36 * 5866 milburn g j and walls d f 1988 _ phys . rev . a _ * 38 * 1087 marian p and marian t a 1996 _ j. phys . a : math . gen . _ * 29 * 6233 horodecki r , horodecki p , horodecki m , and horodecki k 2009 _ rev . phys . _ * 81 * 865 modi k , brodutch a , cable h , paterek t , and vedral v 2012 _ rev phys . _ * 84 * 1655 adesso g , ragy s , lee a r 2014 _ open syst . * 21 * 1440001 ollivier h and zurek w h 2001 _ phys . lett . _ * 88 * 017901 henderson l and vedral v 2001 _ j. phys . a _ * 34 * 6899 yu t and eberly j h 2002 _ phys . rev . b _ * 66 * 193306 maziero j , cleri l c , serra r m , and vedral v 2009 _ phys . rev . a _ * 80 * 044102 mazzola l , piilo j and maniscalco s 2010 _ phys . lett . _ * 104 * 200401 streltsov a , kampermann h , and bruss d 2011 _ phys . lett . _ * 107 * 170502 ciccarello f and giovannetti v 2012 _ phys . rev . a _ * 85 * 022108 serafini a , illuminati f , paris m g a , and de siena s 2004 _ phys . a _ * 69 * 022318 goyal s k and ghosh s 2010 _ phys . * 82 * 042337 souza l a m , drumond r c , nemes m c , and fonseca romero k m 2012 _ opt . commun . _ * 285 * 4453 buono d , nocerino g , porzio a , solimeno s 2012 _ phys . rev . a _ * 86 * 042308 barbosa f a s , de faria a j , coelho a s , cassemiro k n , villar a s , nussenzveig p , and martinelli m 2011 _ phys . rev . a _ * 84 * 052330 isar a 2012 _ phys . _ * t147 * 14015 madsen l s , berni a , lassen m , and andersen u l 2012 _ phys . lett . _ * 109 * 030402 marian p , marian t a , and scutaru h 2001 _ j. phys . a : math . gen . _ * 34 * 6969 marian p , marian t a , and scutaru h 2003 _ phys . a _ * 68 * 062309 pirandola s , spedalieri g , braunstein s l , cerf n j , and lloyd s 2014 _ phys . _ * 113 * 140405 simon r 2000 _ phys . lett . _ * 84 * 2726 peres a 1996 _ phys . lett . _ * 77 * 1413 giorda p and paris m g a 2010 _ phys . rev . * 020503 adesso g and datta a 2010 phys . rev . 105 * 030501 olivares s and paris m g a 2012 _ int b _ * 27 * 1345024 marian p and marian t a 2008 _ phys . * 101 * 220403 williamson j 1936 _ amer . j. math . _ * 58 * 141 marian p , ghiu i , and marian t a 2013 _ phys . a _ * 88 * 012316 bennett c h , divincenzo d p , smolin j a , and wootters w k 1996 _ phys . rev . a _ * 54 * 3824	in this paper we examine the decay of quantum correlations for the radiation field in a two - mode squeezed thermal state in contact with local thermal reservoirs . two measures of the evolving quantum correlations are compared : the entanglement of formation and the quantum discord . we derive analytic expressions of the entanglement - death time in two special cases : when the reservoirs for each mode are identical , as well as when a single reservoir acts on the first mode only . in the latter configuration , we show that all the pure gaussian states lose their entanglement at the same time determined solely by the field - reservoir coupling . also investigated is the evolution of the gaussian quantum discord for the same choices of thermal baths . we notice that the discord can increase in time above its initial value in a special situation , namely , when it is defined by local measurements on the attenuated mode and the input state is mixed . this enhancement of discord is stronger for zero - temperature reservoirs and increases with the input degree of mixing .
neutrinoless double beta decay occurs whenever ordinary beta decay is forbidden due to energy conservation or retarded due to angular momentum mismatch , but the nucleus , which is two units of charge away , is accessible via a order week interaction@xcite-@xcite . it is a process known for almost 70 years , which has been searched for , but not seen yet . it is still of great theoretical and experimental interest since perhaps it is the only process , which can unambiguously determine whether massive neutrinos can be majorana or dirac type particles . it can occur only if the mass eigenstates are majorana particles ( the particle coincides with its antiparticle ) . in the weak basis @xmath0 , @xmath1 the neutrino mass matrix takes the form : @xmath2 + where @xmath3 is a @xmath4 left - handed ( isotriplet ) majorana mass matrix , @xmath5 is a dirac @xmath6 mass matrix and @xmath7 is the right - handed ( isosinglet ) @xmath8 majorana mass matrix . in models motivated by gut s @xmath9 , @xmath10 , @xmath5 is of the order of the up quark mass matrix and @xmath7 is very heavy ( @xmath11 gev ) so that light neutrinos with mass of order @xmath12 can occur . on the other hand in r - parity violating supersymmetry only @xmath3 is non zero . anyway after diagonalizing the above matrix the weak eigenstates are related to the mass eigenstates via the equation : @xmath13 + in the above notation @xmath14 is the usual maki , nakagawa , sakata charged leptonic matrix , which for convenience sometimes will merely be indicated by @xmath15 . one encounters two lepton violating parameters @xmath16 and @xmath17 given by @xcite : _ = < m_>/m_e [ lmasse1a ] < m _ > = ^3_j ( u^(11)_ej)^2 _ j m_j , [ lmasse1 ] ^l__n = ^3_j ( u^(12)_ej)^2 _ j [ lmasse ] with @xmath18 ( @xmath19 ) the light ( heavy ) neutrino masses and @xmath20 ( @xmath21 ) the majorana phases of the corresponding mass eigenstates . the separation of these particle physics parameters from the nuclear physics holds , if the neutrino is very light ( @xmath22 ) or much heavier than the proton ( @xmath23 ) . in the presence of right - handed currents one can have a contribution via light intermediate neutrinos , in which the corresponding amplitude does not explicitly depend on the neutrino mass @xcite(see fig . [ lnomass ] ) . we encounter two lepton violating parameters : in the presence of right - handed currents one can have additional mass dependent terms ( see [ hmass ] ) . the corresponding lepton violating parameter is given by : ^r__n = ( ^2 + ^2+c_0 ) ^3_j ( u^22_ej)^2 _ j [ rmasse ] -\psi_2 ) ] \nonumber\\ & & + \|x_r\|^2 + re ( { \tilde c}_{0\lambda } \lambda x_{r } + { \tilde c}_{0\eta } \eta x_{r } ) ] , \label{eq:70 } \end{aligned}\ ] ] where @xmath28 @xmath29 @xmath30 where the subscript @xmath31 indicates heavy particle ( neutrino ) , @xmath32 is the relative phase between @xmath33 and @xmath34 and @xmath35 is the relative phase between @xmath33 and @xmath36 , while @xmath37 depend on the energy and nuclear physics the ellipses \ { ... } indicate contributions arising from other particles such as intermediate susy particles , unusual particles predicted by superstring models , unusual messengers in extra dimensions , exotic higgs scalars etc . 9 j.d . vergados , phys . rep . * 361 * , ( 2001 ) 1 j.d . vergados , phys . rep . * 133 * ( 1986 ) 1 ; t. tomoda , rep . * 54 * ( 1991 ) 53 ; j. suhonen and o. civitarese , phys . * 300 * ( 1998 ) 123 ; a. faessler and f. imkovic , p. vogel , double beta decay : nucl - th/0005020 ; nucl - th/9904065 . bilenky , amand faessler and f. simkovic , phys . d * 70 * ( 2004 ) 033003 ; hep - ph/0402250 . g. przeau , m. ramsey - musolf and p. vogel , phys . d * 68 * ( 2003 ) 034016 . g. pantis , f. imkovic , j.d . vergados , and a. faessler , phys . c * 53 * , 695 ( 1996 ) . a. faessler , s. kovalenko , f. imkovic , and j. schwieger , phys . lett . * 78 * , 183 ( 1997 ) . you ke _ et al _ , b * 265 * ( 1991 ) 53 . h.v . klapdor - kleingrothaus _ et al _ , eur . j. * 12 * ( 2001 ) 147 ; c.e . et al _ , phys . rev . c * 59 * ( 1999 ) 2108 . et al _ , phys . c * 63 * ( 2001 ) 065501 . et al _ , phyds . c * 63 * ( 2001 ) 065501 f.a . et al _ , phys . c * 62 * ( 2000 ) 044501 . t. bernatowicz _ et al _ , phys . c * 47 * ( 1993 ) 806 . a. allesandrello _ et al _ , phys . b * 486 * ( 2003 ) 13 . r. luesscher _ et al _ , phys . b * 434 * ( 1998 ) 407 . a de silva _ et al _ , phys . c * 56 * ( 1997 ) 2451 . klapdor - kleingrothaus _ et al _ , phys . lett.b * 578 * ( 2004 ) 54 . m. gozdz , w. a kaminski ana a. wodecki , phys . c * 69 * ( 2004 ) 025501 g. battacharrya , h.v . klapdor - kleingrothaus , h. paes and a. pilaftsis , phys . rev d * 67 * ( 2003 ) 113001 .	neutrinoless double beta decay pops up almost in any extension of the standard model . it is perhaps the only process , which can unambiguously determine whether the massive neutrinos are majorana or dirac type particles . in addition from the lifetime of this decay , combined with sufficient knowledge of the relevant nuclear matrix elements , one can set a constraint involving the neutrino masses . furthemore , if one incorporates the recent results of the neutrino oscillation experiments , one can determine or set a stringent limit on the neutrino mass scale . in addition one may obtain usefull information regarding the presence of right handed currents and the right handed neutrino mass scale . one can also constrain the parameters of supersymmetry and , in particular , set limits in of r - parity violating couplings as well as get information about extra dimensions .
the quantum mechanical regime of gravitation continues to be opaque to our probing , both theoretically and experimentally , in the latter case due to the extreme weakness of the gravitational interaction . there are now many theoretical probes , including ( but not limited to ) string theory , loop quantum gravity , spin foams , causal sets , causal dynamical triangulations and asymptotic safety . however , none of these are yet complete or able to predict phenomena which can be seen experimentally . given the lack of experimental direction , we must rely heavily on philosophy and mathematics to justify any particular approach . these circumstances have motivated the careful study of mathematical toy models , often obtained by imposing symmetry and/or changing the number of dimensions of spacetime . such models , if exactly solvable , could help us to see around the obstacles occluding our view of the quantum regime of realistic theories of gravity . it is also possible that results obtained in the toy models might survive the transition to realistic models , and provide the basis for attempts to experimentally test quantum gravity . gravitational collapse of a spherically symmetric thin shell of dust in 3 + 1 dimensions presents a simple model of a matter gravity interaction that has both classical and quantum solutions . classically , the theory is parametrized completely by a pair of phase space variables : the radius @xmath1 of the shell and its conjugate momentum @xmath2 . this is effectively a theory of particle mechanics , yet it is rich enough to describe dynamical black hole formation , including the evolution of apparent horizons and singularity formation . on the quantum side , this is a rare solvable model which describe a fully quantum interaction between gravity and matter . as such significant attention has been given to the quantum theory@xcite . perhaps the most influential work in this area is that of hjek , kay and kucha @xcite where a wheeler - dewitt quantization was successfully completed for shells with a rest mass @xmath3 ( in planck units ) . using a particular choice of time coordinate @xmath4 and a separation ansatz @xmath5 , the authors were able to solve for the complete spectrum of time - independent bound and scattering states . in the present paper , we study quantization of the thin shell model along the same line as hjek _ et al_. the key difference is that we employ polymer quantization @xcite rather than standard techniques which assume that spacetime is a continuum . polymer quantization posits that space is fundamentally discrete near or below some microscopic scale @xmath6 , and that the continuum emerges only as a coarse - grained approximation . our results agree with hjek _ et al _ in the @xmath7 limit , and we find interesting new results for the case of small but finite @xmath6 . in particular we are able to consistently quantize the theory for rest mass @xmath8 , and using numerical methods we are also able to solve the fully quantum equations of motion . in this section we establish the classical theory upon which the quantum theory is built . we begin by sketching a derivation of the classical equations of motion for a thin shell of dust collapsing under its own gravitation . details can be found , for example , in chapter three of eric poisson s text book @xcite . after these equations of motion are obtained we develop the canonical theory to set up quantization in the next section . now consider the spacetime corresponding to a collapsing thin shell , which represents a codimension - one hypersurface of spacetime . outside of the shell the spacetime is described by the schwarzschild metric , while inside the shell spacetime is flat . the ` in ' , out and ` shell ' line elements are : ds^2 _ & = & - dt^2 + dr^2 + r^2 d ^2 + ds^2 _ & = & - f(r ) dt^2 + f^-1(r ) dr^2 + r^2 d ^2 + ds_^2 & = & - d ^2 + r^2 d ^2 where @xmath9 and @xmath10 is the _ total _ gravitational energy of the shell . the next step is to impose the junction conditions at the shell in order to form a complete spacetime . these conditions require that the metric induced on the shell hypersurface @xmath11 from the ` out ' spacetime agrees with the metric induced at the shell from the ` in ' spacetime . furthermore , the junction conditions tell us that a non - trivial stress - energy is associated with the shell , given by [ s1 ] t_ab = ( [ k_ab]-[k]h_ab ) , where square brackets denote the difference between the outer and inner values of the extrinsic curvature @xmath12 and its trace @xmath13 . we take the stress energy of the shell to be that of a perfect fluid with 2d density @xmath14 travelling with the velocity @xmath15 of the shell [ s2 ] t_ab = u_a u_b . equating the stress energy given in ( [ s1 ] ) with that of ( [ s2 ] ) is most easily done in terms of ` shell ' coordinates ( @xmath16 ) . doing so reveals that : 1 ) the rest mass @xmath17 of the shell is a constant of motion ; 2 ) the first integral of the radial equation of motion is e= m-=. [ eq : m1 ] the velocity in the above equation is determined using the time coordinate @xmath18 of a co - moving observer on the shell . this implies a foliation of the spacetime into hypersurfaces labelled by constant values of @xmath18 . going to a canonical theory in terms of @xmath1 and its conjugate momentum @xmath2 , one finds that the hamiltonian is not easily represented as an operator in the quantum theory . following the steps outlined below , one arrives at the following hamiltonian : h = m - . in the quantum theory , the @xmath19 term must be treated as an infinite sum which is difficult to work with , although significant progress was made by hajicek@xcite . if we instead choose the inner flat space time coordinate @xmath4 , we obtain a hamiltonian that is more manageable . the relationship between the two time coordinates in question is @xmath20 , from which one can derive ( ) ^2 = . putting this into ( [ eq : m1 ] ) gives : [ basic ] e = - . where from now on a dot represents a derivative with respect to the inner time coordinate @xmath4 . this is the energy of a relativistic particle in a coloumb potential @xmath21 . in order to arrive at the canonical theory , we take the total energy to be the hamiltonian , that is , @xmath22 , and find the momentum which is conjugate to @xmath1 for this hamiltonian : p = = . from this we obtain the following expression : [ htilde ] ( h + v)^2 = p^2+m^2 . at this point we could take the phase space parametrized by @xmath23 , solve for the hamiltonian @xmath24 in terms of these variables and proceed with quantization . however , the resulting hamiltonian has a square root which is difficult to handle in the quantum theory . to avoid this problem we proceed by putting the hamiltonian into parametrized form , as in @xcite , by extending the phase space to include the time @xmath4 and its momentum @xmath25 . the equation ( [ htilde ] ) is treated as a constraint as described in the remainder of this section . the extended phase space has two canonical pairs @xmath23 and @xmath26 where the momentum conjugate to time is @xmath27 . the canonical action is i = p + p - h , where @xmath28 is the evolution parameter and @xmath29 is a lagrange multiplier . the dynamics is generated by the hamiltonian constraint [ sh ] h:= -(p - v)^2+p^2+m^2 , which is quadratic in the momenta and does not have a square root . the equations of motion are : [ eom1 ] & = & -2(p - v ) ; + [ eom2 ] & = & 0 ; + [ eom3 ] & = & 2p ; + [ eom4 ] & = & . these expressions are equivalent to the dynamics described by ( [ basic ] ) , which can be seen as follows . the second equation tells us that @xmath30 is a constant of motion . using the first equation , we find from ( [ eom3 ] ) and ( [ eom4 ] ) respectively that [ eoma ] & = & = , + [ eomb ] & = & = - . taking the @xmath4-derivative of ( [ eoma ] ) and combining this with ( [ eomb ] ) , one obtains an expression which is equivalent to the @xmath4-derivative of ( [ basic ] ) . this establishes the classical , canonical theory that we shall quantize in the next section in this section we establish a hilbert space of functions @xmath31 , and represent the phase space variables @xmath32 as operators on this space . we then write ( [ sh ] ) as a quantum constraint @xmath33 in terms of these operators and look for solutions . the polymer hilbert space is built upon the real line with discrete topology ( i.e. the bohr compactification of the real numbers @xmath34 ) @xcite . position eigenstates are defined as : _ x(r ) = \ { ll 1 , & r = x + 0 , & r x . . the basis is uncountable and therefore unseparable , but we will find that solutions to the constraint @xmath33 are defined on a superselection sector which has a countable basis and is separable . the polymer inner product is : ( _ x , _ x ) = _ x , x , where @xmath35 is the kronecker delta . a general , time - dependent state is written as : ( t , r ) = _ x c_x(t ) _ x(r ) , where each @xmath36 is a time - dependent coefficient for the basis state @xmath37 . using the inner product we find that the norm squared is : ( \|(t , r ) , ( t , r ) ) = _ x \|c_x(t ) c_x(t ) , which notably depends upon time . we choose the usual representation for the first pair of operators where @xmath38 acts by multiplication and @xmath39 acts as a partial derivative = t , = i _ x _ x _ x the second pair of operators requires more care . since we have position eigenstates , the action of the position operator is simply _ x = x _ x. however , the usual definition of the momentum operator @xmath40 as a continuum partial derivative is not well defined because space is discrete . in order to construct a well defined @xmath40 , we introduce a finite translation operator @xmath41 which shifts a position eigenstate by @xmath42 : _ _ x(r ) = _ x(r+ ) = _ x-(r ) , a well - defined momentum operator which depends upon the spacing @xmath43 is given by : = ( _ 2 - _ 2^ ) . squaring this yields : ^2 = ( 2 - _ - _ ^ ) . in principle , @xmath6 is free to vary in both time and space . however in order to keep the equations manageable , we choose @xmath6 to be an arbitrary fixed constant . this amounts to partially choosing a lattice . since @xmath6 will enter the constraint equation only in the @xmath44 term , a choice of @xmath6 picks out the superselection sector ; only the coefficients @xmath45 and eigenstates @xmath46 which correspond to points @xmath47 on a regular @xmath6-spaced lattice will have an effect on each other . in general , a lattice is the following set of points l _ = \ { x = + k \| k } , + where the choice of @xmath48 picks the superselection sector . here we study the theory on two of these lattices corresponding to the choices @xmath49 ; one which contains the point @xmath50 and one which does not . notice that the lattices defined above do not have a bound on @xmath47 . this is necessary for the @xmath44 operator to be well - defined ( with the same definition ) on all position eigenstates . and alter the definition of @xmath44 at the first lattice point as done in @xcite ] this means that in principle , we are allowing for negative eigenvalues of the position operator , but since the @xmath51 operator comes from a positive semi - definite radial variable @xmath52 , we will study solutions with support on the @xmath53 lattice points only . there are conserved currents coming from the constraint equation , and we apply boundary conditions which conserve the corresponding charges on the @xmath53 portion of the lattice . we now define the auxiliary hilbert space ( without boundary conditions ) : _ ^:= \ { . ( t , r ) = _ x c_x(t ) _ x(r ) \| _ x l _ c^_x(t ) c_x(t ) } . this is a general space of functions within which the operators and inner product are well - defined . the solutions to @xmath33 will comprise a subset of this space , and appropriate boundary conditions will be introduced to define the charge - conserving portion of this solution space . in this subsection we define the physical hilbert space of solutions to @xmath54 , or equivalently [ opeq ] -^2(t , r ) + ( t , r ) = 0 . using the definitions in the previous subsection , the constraint equation for states in @xmath55 can be written as : _ x = 0 . if we shift the summation for the last two terms , and notice that the above condition must hold independently at each point @xmath47 , we arrive at an equation of motion for the coefficients : [ eom ] _ x - _ x + ( + m^2- ) c_x - ( c_x+ + c_x- ) = 0 . notice that some of the terms in the above equation are divergent at @xmath50 on the lattice with @xmath56 , but only in this case . in order to fix boundary conditions , we first note that there are two conserved charges . these are the polymer analogs to the klein - gordon ( kg ) inner product and the energy form defined in @xcite . given two states @xmath57 and @xmath58 we define the following bilinear forms : [ kg ] q ( , ) & = & _ x ( i ( \|b_x _ x - _ x c_x ) + \|b_x c_x ) , + [ energy ] e ( , ) & = & _ x ( ( _ x _ x + ( \|b_x+ - \|b_x ) ( c_x+ - c_x ) + ( m^2 - ) \|b_x c_x ) ) , where @xmath59 is the kg inner product and @xmath60 is the energy . one can check that @xmath60 and @xmath59 are conserved on the @xmath53 portion of the lattice under the dynamics defined by ( [ eom ] ) provided that one of the following boundary conditions holds : [ c1 ] : & & b_= c_= 0 ; + [ c2 ] : & & b_= b_- , c_= c_- . having a choice of boundary conditions which conserve the charges is different than in the continuum theory where only a single option is available . this is a direct result of discretization : the continuum boundary condition defines the limiting form of the wavefunction at small @xmath1 , while the polymer boundary conditions involve the value of @xmath31 at one or two distinct points using ( [ c1 ] ) or ( [ c2 ] ) respectively . we find numerically ( as described below ) that in the limit of @xmath7 , the two polymer boundary conditions agree @xmath61 , so that there is no contradiction with the continuum theory . we now define the hilbert space of solutions to the quantum constraint by applying additional conditions to the more general @xmath62 : _ i^= \ { . _ ^ \| = 0 , _ i = 0 } , where the subscript @xmath63 labels the choice of boundary conditions : _ 1:=c _ , _ 2:= c_- c_- . states in this hilbert space satisfy the quantum constraint and preserve the @xmath59 and @xmath60 charges on the @xmath53 portion of the lattice . with this definition we can move forward and solve for these states explicitly . in this section we find solutions to the constraint @xmath64 . we first look at time - independent solutions and then study the dynamics . let us now find explicitly the states @xmath65 which are eigenstates of the hamiltonian @xmath66 ( where @xmath67 in this representation ) , within the solution space of the quantum constraint @xmath64 . these are found using the following ansatz for the time - dependent part of the states : [ ansatz ] c_x(t ) = c_x e^-iet , where each @xmath68 is independent of @xmath4 . putting this ansatz in ( [ eom ] ) results in the following difference equation for the constants : ( 2 - ^2 - f_k ) c_k - c_k-1 - c_k+1 = 0 , [ eq : recurs1 ] where we have used the integer label @xmath69 for points on the lattice and defined e^2 - m^2 , f_k ^2 m^2 ( + ) . on a positive energy eigenstate of the form ( [ ansatz ] ) , the kg inner product ( [ kg ] ) and energy form ( [ energy ] ) are q ( , ) & = & _ x ( e+ ) \|c_x\|^2 , + e ( , ) & = & _ x . notice that each summand of @xmath59 is positive semi - definite for @xmath70 . the energy summands may become negative at small values of @xmath47 , but are positive definite for @xmath71 . in the continuum theory , the fall - off conditions at small @xmath1 guarantee that the energy form is positive definite . note however that only the total sums @xmath59 and @xmath60 are physical observables while the summands themselves are not . using ( [ eq : recurs1 ] ) and either boundary condition , the energy form is found to be e ( , ) = e q ( , ) , so that if the states are normalized to make @xmath72 , then the bilinear form @xmath60 measures energy . the difference equation can not be solved analytically , but using the fact that @xmath73 , we can solve the equation at large @xmath74 : [ largek ] c_k = a^k , a 1 - + . these large @xmath74 solutions have different behaviours for different values of energy at fixed @xmath75 and @xmath6 , and this characterizes the states . bound states : : for @xmath76 we have that @xmath77 . normalizable states require the negative choice of sign @xmath78 , giving states which go to zero at large @xmath74 . scattering states : : for energies in the range @xmath79 we get that @xmath80 so that these solutions are neither normalizable nor divergent . energy may be of either sign , and the bounds go to @xmath81 in the limit of @xmath7 . non - existent states : : for the range @xmath82 we have that @xmath83 which requires a positive choice of sign @xmath84 for normalizable solutions . these states have no counterpart in the continuum theory since in the @xmath7 limit , bound and scattering states cover the entire energy spectrum . in the polymer theory , we do not find any states at these energies which satisfy either boundary condition . since the difference equation can not be solved analytically we resort to numerical techniques . a time independent solution is defined by knowing all of the non - zero coefficients @xmath85 , and these can be found using a shooting method . to do so , one checks many different values of @xmath10 to search for those which satisfy the boundary conditions . given a particular ` guess ' @xmath10 , one begins with the solution ( [ largek ] ) at some @xmath74 which satisfies @xmath86 and @xmath87 , then iterates downward according to the full difference equation ( [ eq : recurs1 ] ) . after the coefficients are found all the way down to the first lattice point , one checks whether one of the boundary conditions ( [ c1 ] , [ c2 ] ) is satisfied . if the ` guess ' at @xmath10 results in a solution which satisfies the choice of boundary condition , this state is recorded as an eigenstate . otherwise it is discarded as lying outside the space of solutions . ( all of the results in this section were confirmed using an alternative numerical method based on continued fractions . ) as f_k c_k - c_k-1 = c_k+1 , where @xmath88 . then defining @xmath89 we have the recursion relation _ k-1 = ( f_k - _ k)^-1 . now , at large @xmath74 where the asymptotic solution ( [ largek ] ) holds , we have @xmath90 . iterating downward from here , one can check whether : @xmath91 for the boundary condition @xmath92 ; @xmath93 for the boundary condition @xmath94 . ] ) let us first review the results of the continuum theory for comparison where bound states were found only for @xmath3 ; there are no continuum solutions for @xmath95 which satisfy the boundary conditions . the spectrum for @xmath3 is [ conspec ] e_n = m , = + , where @xmath96 is an integer greater than or equal to zero . notably , the entire spectrum is positive . in the polymer theory we also find bound states , but without any restriction on @xmath75 . these exist on both choices of lattice ( @xmath56 and @xmath97 ) and for both choices of boundary conditions . for @xmath3 the continuum spectrum is recovered in the limit of @xmath7 . in fig . [ ef ] the densities associated to the kg and energy forms are plotted for the lowest four bound eigenfunctions for @xmath98 satisfying @xmath99 on the lattice with @xmath56 and @xmath100 . since the energies are all positive , the kg inner product is positive semi - definite and may be interpreted as the probability density . the two densities have a similar profiles outside of the first ten or so lattice points ; as the eigenvalue increases , both densities move further from the origin and become more spread . increasing @xmath75 tends to focus the densities closer to @xmath0 , while the solutions are insensitive to the choice of @xmath6 . a large @xmath6 gives a coarse approximation to the same curves . for comparison , we also provide plots in fig . [ ef2 ] of the lowest four bound eigenfunctions for @xmath101 satisfying @xmath102 on the @xmath103 lattice . using the probability density , we can calculate radial expectation values and check where stationary states lie in relation to their schwarzschild radii . we note here that such values are not physical observables since @xmath51 does not commute with @xmath33 , and will comment more on this in the discussion . in table [ rad ] the radial expectation values corresponding to the plots in fig . [ ef ] are compared with the schwarzschild radii . none of these expectation values are within the corresponding schwarzschild radius , getting further out as the eigenvalue increases . the radial expectation values lie within the schwarzschild radius only for @xmath104 @xcite , but we are unable to probe low enough values due to numerical limitations . .[rad ] radial expectation values @xmath105 of lowest four energy eigenstates compared with the schwarzschild radii @xmath106 . here we used @xmath98 and the boundary condition @xmath99 on the lattice with @xmath56 and @xmath100 . [ cols=">,^,^,^,^",options="header " , ] figure [ m=0.9 ] shows the spectrum for @xmath107 at various spacings @xmath6 . it is interesting that for the boundary condition ( [ c2 ] ) the ground state may have a negative energy eigenvalue for a certain range of @xmath6 . as @xmath6 goes to zero , this negative eigenvalue goes below @xmath108 and leaves the spectrum , and the next lowest eigenstate becomes the ground state . when @xmath3 , for any choice of lattice or boundary condition the polymer spectrum agrees with the continuum spectrum in the limit of @xmath7 . note that one can not use boundary condition ( [ c1 ] ) on the @xmath109 lattice due to divergences in the difference equation which prevent finding @xmath110 . + the polymer boundary conditions have no dependence on @xmath75 , and so has bound state eigenvalues for @xmath8 which have no counterpart in the continuum theory . as the spacing @xmath7 , the eigenvalues in the @xmath8 spectrum continue to fall for smaller and smaller @xmath6 , and do not converge to any fixed values , so that there is no contradiction between the polymer and continuum theories . [ m=2 ] which shows the spectrum for @xmath111 at various spacings @xmath6 . for @xmath8 the ground state may have negative energy with either boundary condition , at certain values of @xmath6 , but the first excited state and above are found to be positive in all cases tested numerically . + in the continuum theory , scattering solutions are found only when @xmath3 as is the case for bound states . the solutions are given by coulomb wave functions and have a continuous energy spectrum for both positive and negative energies @xmath112 . from the limiting form of the solution at large @xmath74 ( [ largek ] ) , scattering solutions in the polymer theory are c_k ~e^i k where @xmath113 . at large @xmath74 these solutions are ingoing and outgoing modes . on the lattice with @xmath56 and using boundary condition ( [ c1 ] ) , one finds that the coefficient @xmath114 obtained from the shooting method starting with @xmath115 at large @xmath74 , is the complex conjugate of the coefficient @xmath116 obtained from the solution which is @xmath117 at large @xmath74 , e.g. @xmath118 for any choice of @xmath10 in the scattering range . [ scatter ] for an example . therefore the following linear combination satisfies ( [ c1 ] ) for any scattering energy : _ e = ^+ - e^2i c_0^+ ^- , where @xmath119 are the solutions with @xmath120 at large @xmath74 . a similar argument applies for other choices of lattice and boundary condition ( [ c1 ] , [ c2 ] ) . this argument shows that the scattering spectrum is continuous for either choice of boundary condition and any choice of @xmath121 , as one would expect . furthermore , at large @xmath74 the scattering solutions are a linear combination of ingoing and outgoing modes indicating a bounce at the origin with a particular phase shift as in the continuum theory @xcite . in the previous subsection we obtained the complete energy spectrum , and each value in the spectrum labels a solution to @xmath122 . consider a hilbert space defined using the kg inner product on such states . if we restrict to the positive energy states , the kg density is positive semi - definite and may be used as a measure of probability . in this hilbert space , an arbitrary state is written as = _ e _ e(t ) _ e. in principle , one could build an electronic database which represents the hilbert space described in the previous paragraph . this would involve storing the coefficients of a large number of energy eigenstates for a fixed @xmath75 , on a given lattice and for a given choice of boundary condition . one could then attempt to construct semiclassical initial data and study the evolution of such data by calculating the equations of motion for the coefficients @xmath123 . however , we have not taken up this task here . rather , we pursue a numerical solution to ( [ eom ] ) based on the finite differences naturally defined by the polymer theory . from this angle , we can ask whether the dynamics we find can be described using positive energy eigenstates only . the equation ( [ eom ] ) is well - suited to numerics since the terms coming from @xmath44 effectively define a second order finite difference derivative with local error terms on the order of @xmath124 . the lattice is also fixed by the polymer theory , up to choices of @xmath121 and @xmath6 , and the boundary conditions necessary to conserve the @xmath59 and @xmath60 charges have been given in ( [ c1 ] , [ c2 ] ) . in fact , the only thing left to do is to choose a method of time integration . for this purpose we use heun s method , which is second order accurate in the time step @xmath125 . stability ( the cfl condition @xcite ) requires that we limit the number of lattice spaces that data can move in a single time step , which is done by setting @xmath126 . in the code , we specify initial coefficients @xmath45 and initial velocities @xmath127 , then integrate forward in time according ( [ eom ] ) . in a collapse problem such as this , one often encounters difficulty near the origin where the data may become large relative to values elsewhere . summing large numbers near @xmath0 can yield significant round - off errors and lead to numerical instability . because of this it is often useful to impose a fall - off condition on the solution in order to improve the stability of the code . in order to improve stability we borrow from the continuum theory , since the numerical solutions converge to continuum solutions in the @xmath7 limit . in @xcite it was found that the fall - off behaviour at small @xmath1 must be @xmath128 where @xmath129 . there exists another solution in the continuum theory which falls - off as @xmath130 , but this one does not satisfy the boundary condition . for @xmath3 the ` good ' solution goes to zero while the ` bad ' solution diverges , but for @xmath131 both solutions have the same fall - off up to oscillatory terms . for @xmath3 , we improve the stability of our code by writing the coefficients as @xmath132 and solving for the @xmath133 data at each time step . we monitor stability by tracking the charges @xmath59 and @xmath60 throughout the dynamics . decreasing the lattice spacing and time step improves the conservation while increasing run - time . we study semiclassical ingoing initial data of the following form : c_x(0 ) ~e^- , _ x(0 ) ~ , with the proportionality constants chosen to fix @xmath72 . for such data , the probability density is an inward - moving ( up to some small outgoing corrections ) gaussian pulse of width @xmath134 centred at @xmath135 , and the kg density is positive semi - definite . we choose the proportionality constants to normalize @xmath136 so that @xmath60 is a measure of the energy . in the following cases , we study the dynamical behaviour for different values of the rest mass @xmath75 on the lattice with @xmath56 . animations of the kg and energy densities for each of the solutions discussed below can be found at http://ion.uwinnipeg.ca/~gkunstat/polymer/[http://ion.uwinnipeg.ca/ gkunstat / polymer/ ] . @xmath137 : : for small values of rest mass we find that the wavefunction collapses inward , bounces at the origin , then moves outward toward infinity . animations of the kg and energy densities for @xmath138 with @xmath139 can be found at the web address given above . the energy density shows two peaks which bounce off the origin while maintaining their shape . the kg density begins as a single pulse which splits during the bounce leaving a completely negative region which coincides with the outer energy peak . due to this negative region , we are hesitant to refer to the kg density as a probability density . the total energy is @xmath140 in planck units with deviations of less than @xmath141 , while @xmath59 remains fixed at @xmath142 with deviations of less than @xmath143 . @xmath144 : : as the rest mass approaches the planck mass from below , we find we find that the wavefunction continues to bounce off the origin , but now exhibits self - interaction ( dispersion ) throughout the process . the increasing spread of the densities implies the state is moving away from semiclassicality and becoming more quantum as @xmath75 increases . as an example , we have generated animations of the kg and energy densities for @xmath145 . here the total energy is @xmath146 with deviations of less than @xmath147 , while @xmath59 remains normalized within @xmath148 . @xmath131 : : the numerics become increasingly unstable as @xmath75 exceeds the planck mass , and may crash for certain choices of the other parameters . given that this polymer theory is effectively a finite difference approximation of the continuum , we expect the numerical solution to follow the continuum behaviour more closely as @xmath7 . since the continuum theory does not have solutions for this range of @xmath75 , we expect the numerics to fail for small @xmath6 . these numerical difficulties are lessened for large values of @xmath6 which moves the first lattice point further from the origin where the potential diverges . has a similar effect . ] for example , with @xmath149 and @xmath150 , the code becomes unstable when the pulse reaches the origin , while it exhibits a bounce ( with some noise near the origin ) when @xmath100 . see for example the animations of kg and energy densities for @xmath149 with @xmath100 . here the total energy is @xmath151 but undergoes a large deviation of @xmath152 during the bounce , while the kg inner product remains conserved within @xmath153 . for all parameter ranges for which the code is stable , we find generically that the probability density becomes negative locally at some time during the evolution , and these negative portions do not go away as evolution continues . it is possible that such an effect could be due to numerical inaccuracy . however , negative kg density for ingoing gaussian initial data is seen even for very small values of @xmath75 where the numerics are optimal . the implication is that the dynamics we find numerically can not be described in terms of positive energy eigenstates , i.e. these states @xmath31 are not within the hilbert space discussed at the beginning of the section . whether negative eigenstates are a requisite component of any bouncing state is an open question . we studied the polymer quantum mechanics of a self - gravitating thin shell . the energy spectrum was found by solving for stationary eigenstates using the shooting method . when @xmath154 , the spectrum matches the continuum spectrum found in @xcite as the polymer scale @xmath7 . we find two independent choice of boundary conditions which conserve energy and the kg inner product , and in the limit @xmath7 these are both satisfied and are consistent with the ( @xmath3 ) continuum boundary condition . for finite values of @xmath6 we find that there may be negative energies associated with bound states which is something not seen in the continuum theory . for @xmath131 where no continuum solutions exist , we find that the bound state spectrum does not converge to fixed values as @xmath7 , so that there is no contradiction . in the classical theory , all initial conditions lead to black hole formation . bound states are not present , so it is very interesting to find them in the quantum theory . replacing the classical particle ( shell ) by a quantum wavefunction has allowed for a superposition of ingoing and outgoing solutions with total energy @xmath155 which satisfy the boundary condition . for positive energies , these states have positive definite kg density . if we interpret this as the probability density , these are highly quantum states with a large uncertainty in position . scattering states for @xmath156 are explicitly a combination of ingoing and outgoing states at large @xmath47 . since charges are conserved by these states , no energy is lost down a singularity at @xmath0 , indicating a unitary quantum evolution . a question this work brings to light is the meaning of states in quantum gravity . each state is a functions of both time and space , representing a superposition of shell spacetimes . in this formalism , how can we extract semiclassical information and recover a single notion of spacetime ? the physical observables seem limited to the total energy , which defines the location of a schwarzschild radius but does not indicate the shell location . if one wants to learn more about the spacetime , it seems something more is needed . to this end we considered the radial expectation value to represent the position of the shell . even classically , the value of the radius at a particular time is a slicing ( gauge ) dependent quantity . in quantum gravity we may need to be open to such gauge - dependent observables . we studied the dynamics of an ingoing gaussian probability density , taken to represent a semiclassical ingoing shell for small @xmath157 , and becoming increasingly quantum ( with a larger spread in probability density ) as @xmath75 approaches the planck mass from below . the probability density is positive semi - definite only for positive energy eigenstates . for all @xmath75 less than one planck mass , we find that the wavepacket bounces off the origin , but picks up regions of negative probability density . this implies that the bouncing states we find can not be represented as a sum over positive energy eigenstates . this suggests that negative energy eigenstates may play an important role in quantum singularity resolution . it is not clear whether or not the standard physical interpretation of these states as positive energy solutions moving background in time is relevant in the present context . in any case , the present model seems to be an ideal theoretical laboratory for studying this and other important issues in quantum gravity . t. vahaspati , d. stojkovic , quantum radiation from quantum gravitational collapse , phys . b 663 * 107 ( 2008 ) , arxiv : gr - qc/0701096 ; t. vachaspati , schrdinger picture of quantum gravitational collapse , class . * 26 * 215007 ( 2009 ) , arxiv:0711.0006 [ gr - qc ] .	we study the quantum mechanics of self - gravitating thin shell collapse by solving the polymerized wheeler - dewitt equation . we obtain the energy spectrum and solve the time dependent equation using numerics . in contradistinction to the continuum theory , we are able to consistently quantize the theory for super - planckian black holes , and find two choices of boundary conditions which conserve energy and probability , as opposed to one in the continuum theory . another feature unique to the polymer theory is the existence of negative energy stationary states that disappear from the spectrum as the polymer scale goes to zero . in both theories the probability density is positive semidefinite only for the space of positive energy stationary states . dynamically , we find that an initial gaussian probability density develops regions of negative probability as the wavepacket approaches @xmath0 and bounces . this implies that the bouncing state is a sum of both positive and negative eigenstates .
the importance of taub - nut solutions covers a wide area of application that extends from general relativity to string theory . these solutions @xcite are characterized by non - vanishing nut charges . as a result , they are locally asymptotically flat , i.e. , their boundaries are not @xmath3 but @xmath4 fiber over @xmath5 . these boundaries are topologically interesting since they do have a non - vanishing first chern number @xmath6 which is proportional to the nut charge . in these space - times one can not define a global time function , therefore , these manifolds can not be foliated using constant time surfaces and it is not possible to describe their time evolution through a unitary hamiltonian evolution . taub - nut spaces are also characterized by the existence of zero - dimensional fixed point set of the @xmath7 isometry generated by the time - like killing vector @xmath8 , which is called a nut . this is in contrast with the higher - dimensional fixed - point set appears in spherically symmetric black hole solutions , which is called a bolt . tau - nut solutions also possess a singular one dimensional string analogous to dirac string @xcitewhich is called misner string . the singularity of misner string is a coordinate singularity , since one fails to describe the manifold using a single coordinate system . misner string contributes to the entropy of the black hole , therefore the entropy is not simply one quarter the area of the horizon . this is a very interesting property in gravitation , which reveals some features of the geometric entropy and has been studied by hawking and hunter @xcite . another attractive feature of the taub - nut solution is the celebrated gross - perry - sorkin monopole@xcite . this monopole can be constructed by adding another trivial dimension to get a five dimensional metric followed by kaluza klein compactifing the theory to four - dimensional along the euclidian time direction . as a result , one gets a singular four - dimensional monopole solution . this shows how perfectly regular higher dimensional solutions can give rise to a singular lower dimensional one . taub - nut solution plays an important role in revealing the nature of d6-branes in type iia string theory and its relation to m - theory(for a review see _ e.g._@xcite ) . probing the background geometry of n coincident d6-branes using d2-branes shows that the transverse dimensions to both branes are not three , but in fact four dimensions and the metric is the taub - nut metric . the extra dimension is the euclidian time direction which is compact . it comes from the gauge degrees of freedom of the @xmath7 gauge theory living on the d2-brane world - volume . this shows that the d6-brane is a kaluza - klein monopole from the m - theory prospective . another interesting application of nut charged space - times is to extend and test the validity of the ads / cft correspondence to locally asymptotically ads solutions , and study their thermodynamics@xcite . therefore , it is interesting to find new taub - nut and taub- bolt solutions in higher dimensions with an eye on possible applications of these results in various supergravity / string theories . here we present a class of higher dimensional @xmath7 electrically charged solutions with nut charge , which are generalizations to the solutions found in @xcite ( see also @xcite for discussions on their thermodynamics ) . the action for einstein - maxwell theory in @xmath0dimensions for asymptotically ( anti)de sitter space - times is @xmath9 where @xmath10 is the cosmological constant , @xmath11 is newton s constant in @xmath0dimensions , and @xmath12 , where @xmath13 is the vector potential . varying the action with respect to the metric and the vector field @xmath13 we get the following field equations g_-g_&=&2 t _ , + _ ( f^)&=&0 , where t_=f_f_^-1 4g_f^f_. here we consider only the euclidian sections of these metrics . to go to the lorentzian sections one can analytically continue the coordinate @xmath14 and the parameter @xmath15 ( i.e. , @xmath16 , @xmath17 , @xmath18 and @xmath19 ) . the general form of the taub - nut / bolt metric is given by ds^2=f(r)^-1dr^2+(r^2-n^2)d_b^2+f(r)(d+a)^2where the metric @xmath20 is over an even dimensional einstein - khler manifold @xmath21 , @xmath22 is some function of @xmath23 . @xmath15 is called the `` nut charge '' and @xmath24 is the potential of the khler form @xmath25 = d a the hypersurface defined by @xmath26 is a @xmath7 fiber over the base manifold @xmath21 . the base space @xmath21 can be any even dimensional einstein - khler manifold . the case when @xmath27 is special , since the solution obtained is a non - singular space . any other base space will be singular and therefore it can not be considered as a gravitational instanton . on the other hand these solutions are fine when they describe bolt solutions , since these singularities will be hidden behind horizons . in the @xmath28 case the resulting hypersurfaces are squashed @xmath29 and @xmath15 here measures the amount of squashing . in order to generalize this construction by giving electric charges to these solutions , we need some ansatz for the one - form potential @xmath30 . we are going to adopt the following ansatz for the potential a = g(r)(d+a ) the general solution for @xmath31 depends on two parameters , one is the electric charge , @xmath1 and the other is @xmath2 , which can be viewed as potential at infinity ; as @xmath32 we have a=(q / r^d-3+v)(d+a ) . @xmath2 plays an important role in studying the thermodynamics of riessner - nordstrom ( a)ds solutions as has been revealed in @xcite . requiring regularity of the one - form potential @xmath30 at the horizon relates the two parameters @xmath2 and @xmath1 . in fact we will not discuss the de sitter versions of these solutions , but they can be obtained , simply by taking @xmath33 which will be left for future work . while writing this article the work of @xcite has been posted . the authors of @xcite discuss similar solutions to the one presented here , but one can spot two general differences ; i ) the solutions presented here are a bit more general since it has two extra parameters @xmath2 and @xmath1 , ii ) here we also discussed the conditions of having taub - nut or taub - bolt solutions and we have showed the possibility of having a taub - nut for certain values of the mass parameter and @xmath2 . this article is organized as follows ; in the first section we discuss conditions for obtaining nut and bolt solutions for asymptotically locally flat or locally ads space - times in four dimensions . in the following section we present our results in six dimensions which can be generalized to eight and ten dimensions for asymptotically flat or ads space - times . finally we put some final remarks on the relevance of these solutions . the first electrically charged four - dimensional taub - nut solution has been introduced in @xcite . more general versions of this solution have been studied in e.g. , @xcite . their supersymmetry has been discussed in @xcite , where the authors have showed that these solutions do preserve some supersymmetry for certain choices of their parameters . here we present an ( a)ds version of this solution with the additional parameter , v , that we mentioned in the introduction . the charged taub - nut ( anti)-de - sitter space version of the brill s solution has the form ds^2=f(r)(d-2nd)^2+dr^2 f(r)+(r^2 -n^2)(d^2+sin^2d^2)where @xmath34 is given by f(r)=r^4+(l^2 - 6n^2)r^2 - 2ml^2r - l^2(q^2 + 4qvnk+4n^2v^2(k^2 - 1)-n^2)-3n^4 l^2(r^2-n^2 ) . + the gauge potential has the form a = g(r)(d+2nd ) . here @xmath31 is given by g(r)=-vr^2+n^2 + 2nkr r^2-n^2-qr r^2-n^2 , + @xmath1 is the electric charge , @xmath35 is the mass parameter , @xmath36 is some constant and @xmath34 upon going to the lorentzian section one must set @xmath37 and @xmath2 is the potential at infinity . in order for this solution to describe a nut , we must have @xmath38 , so that all of the extra dimensions collapse to zero size at the fixed - point set of the killing vector , @xmath8 . this only happens at a specific value of the mass parameter , @xmath35 , and the parameter @xmath1 , namelym_n = n-4n^3/l^2 q = -2nv(1+k ) otherwise the solution will have a horizon radius @xmath39 and become a bolt . amusingly the last condition , i.e. , @xmath40 can be obtained from requiring the regularity of the one - form @xmath30 at the horizon as well . in fact this is might be a sign of consistency for including the extra parameter , @xmath2 . satisfying the above conditions gives @xmath34 the following form f(r)=(r - n)(r+n)l^2[(r - n)(r+3n)+l^2)].notice here that the roots of @xmath41 $ ] are all less than @xmath15 . since the fiber has to close smoothly at @xmath42 , we do that by setting the period @xmath43 of the @xmath14 direction to be @xmath44 . this implies that = 8n notice that the value of the mass parameter that produces the nut in the charged case is identical to that of the uncharged case . furthermore the stress - energy tensor vanishes upon satisfying the above conditions . as a result the metrics for the two cases in four - dimensions are the identical . it will be clear from the discussion below that this is not the case for higher dimensional solutions . in higher dimensions the mass parameter of the charged case will depend on both the nut charge and the electric charge . this is a new feature that makes higher dimensional charged nut solutions more richer and possess one more parameter , namely @xmath1/or @xmath2 , as it is clear from curvature tensor calculation . another feature which is unique to the four - dimensional solution is that the existence of the additional parameter @xmath2 together with the nut charge induces a magnetic charge and changes electric charge as well even at large radius . as we go to the asymptotical flat limit @xmath45 the solution will have no magnetic charge . the magnetic monopole appears as a result of the non - trivial fibration . also in this limit the electric charge will be only @xmath1 . this shows that the parameters @xmath1 and @xmath2 are not convenient for the four - dimensional solution and they should be replaced by @xmath46 and @xmath47 , the total electric charge and the magnetic charge of the solution as in @xcite , which shows that this solution is a dyon . this phenomenon does not happen in higher dimensions , i.e. the @xmath1 will be the total electric charge up to some numerical factor without any contribution from @xmath2 , but it is clear from the electric field calculation that there is a linear charge density term as well which is proportional to @xmath48 . as a result we keep parameterizing these solutions using @xmath1 and @xmath2 in the higher dimensional solutions . to have a regular bolt solution at the horizon , i.e. , at @xmath49 , the following two conditions must be satisfied simultaneously : + ( a ) @xmath50 + ( b ) @xmath51 + the first condition is just the definition of a horizon and the second is following from the fact that we need to avoid conical singularities at the bolt , keeping at the same time the periodicity of @xmath14 to be @xmath52 . in addition to these requirements we have another regularity requirement , namely ; regularity of the one - form potential @xmath30 at the bolt @xmath53 . this leads to the following relation q = -v(r_b^2+n^2 + 2nr_bk)imposing condition ( a ) , the mass parameter is given by m = m_b=1/2[r_b^4+(l^2 - 6n^2)r_b^2 + 4l^2n^2v^2(1-k^2)-3n^4+l^2n^2 - 4qkl^2vn - q^2l^2]l^2r_b by imposing condition ( b ) one gets @xmath54 as the solution of the following fourth order algebraic equation , 6r_b^4n - r_b^3l^2+(2v^2l^2n+2nl^2 - 6n^3)r_b^2 - 2n^3v^2l^2=0 one can choose to write the previous expressions in terms of either @xmath1 or @xmath2 . here we have chosen @xmath2 since it is a bit easier , otherwise the order of the algebraic equation will be higher . one can check numerically that there are solutions ( with @xmath55)to the above algebraic equation in certain ranges of @xmath2 and @xmath15 for a fixed @xmath56 . but unless we get some expression for the entropy and the specific heat we can not estimate the correct ranges of the above parameters and we leave that for a future work . here we are presenting the locally asymptotically flat version of the above solutions . the four dimensional solution has the form ds^2=f(r)(d-2nd)^2+dr^2 f(r)+(r^2 -n^2)(d^2+sin^2d^2)where @xmath34 is given by f(r)=r^2 - 2mr-(q^2 + 4qvnk+4n^2v^2(k^2 - 1)-n^2)(r^2-n^2 ) . + the gauge potential has the form a = g(r)(dt+2nd ) . here @xmath31 is given by g(r)=vr^2+n^2 + 2nkr r^2-n^2+qr r^2-n^2 , + . as before , in order for this solution to describe a nut solution , we must have @xmath38 , so that all of the extra dimensions collapse to zero size . this only happens at specific value of the mass parameter , namely m_n = notherwise the solution will have a horizon radius @xmath57 and will describe a bolt . choosing the mass parameter to have the above value , @xmath34 gets the following form f(r)=(r - n)(r+n ) in order to get the fiber to close smoothly at @xmath42 , one should set the period @xmath43 of the @xmath14 direction to be @xmath58 . this implies that = 8n.as we have mentioned the existence of the parameter @xmath2 together with the nut charge induce a magnetic charge @xmath59 which vanishes in the asymptotic flat limit @xmath45 . in the case of bolt solution , we impose the regularity condition that the one - form potential @xmath30 vanishes at the bolt @xmath53 . it leads to the same relation q = -v(r_b^2+n^2 + 2nr_bk)also , imposing condition ( a ) , the mass parameter is given by m = m_b=1/2[r_b^2 + 4n^2v^2(1-k^2)+n^2 - 4qkvn - q^2 ] by imposing condition ( b ) one gets @xmath54 as the solution of the following fourth order algebraic equation , -r_b^3 + 2n(v^2 + 1)r_b^2 - 2n^3v^2=0 in fact one can solve the above equation analytically but instead of presenting the relatively complicated expression we plotted @xmath54 in [ rootplot ] as a function of @xmath2 , where @xmath60 . one can see clearly that @xmath61 . there are two different choices for the base manifold in six dimensions , namely ; @xmath62 and @xmath63 . let us start with the @xmath62 base manifold . using @xmath62 as a base space , the metric of the charged taub - nut - ads solution has the form @xmath65 where @xmath66 is the metric over @xmath67 which has the following form @xmath68 where f(r ) is given by f(r)= & & -the gauge potential has the form a = g(r)(d+u^2 2(1+u^2/6)(d+d ) ) where @xmath31 is given by g(r)= q + v this solution is a generalization to the solutions found in @xcite . the thermodynamics for the uncharged version of these solutions have been studied in @xcite . for this solution to describe a nut , one must impose @xmath38 , this happens when the mass parameter takes the value m_n= otherwise the solution will have a horizon radius @xmath57 and the solution will describe a bolt . notice here that the value of the mass that causes the nut is not the same as the one for the uncharged solution , as we mentioned in the previous section , since we have a dependence on either @xmath2 or @xmath1 . choosing the mass parameter to have the above value @xmath34 will have the following form f(r)=&&the regularity of the potential @xmath30 at @xmath42 require that q= -8n^3v.again , here one can choose to write the previous expressions in terms of either @xmath1 or @xmath2 . as before we are going to write expressions in terms of @xmath2 . the fiber has to close smoothly at @xmath42 , this can be achieved by setting the period @xmath43 to be = 12 n.in order not to get any solution with @xmath69 which might spoil the nut solution we must demand that the expression in the square brackets in eqn . ( 3.38 ) does not have any positive roots greater than @xmath15 . this can be achieved by setting @xmath70 in this expression which assumes the form & & [ 3x^5 + 30nx^4+(l^2 + 114n^2)x^3+(192n^3 + 8nl^2)x^2 + & & + ( -12n^2l^2v^2 + 120n^4 + 20n^2l^2)x+16l^2n^3 ] , then by requiring @xmath71 can produce a nut ] in this expression , it will have no positive roots . the above requirement might serve as a condition on @xmath2 for obtaining a nut solution in this case . there is no four dimension analog for this condition . as we have mentioned in the introduction for @xmath64 case taub - nut solution have no curvature singularities at @xmath42 , indeed , r^ijklr_ijkl ~ ( r+n)^-12 as in four dimensions , in order to have a regular bolt solution at the horizon , i.e. , at @xmath49 , the following two conditions must be satisfied simultaneously : + ( a ) @xmath50 + ( b ) @xmath72 + the first condition is just the definition of a horizon and the second is following from the fact that we need to avoid conical singularities at the bolt , keeping at the same time the periodicity of @xmath14 to be @xmath73 . in addition to these requirements we have another regularity requirement , namely ; regularity of the one - form potential @xmath30 at the bolt @xmath53 . this leads to the following relation v =- qr_b ( r_b - n)^2imposing condition ( a ) , the mass parameter is given by m = m_b=-1/2(-r_b^4 + 6n^2r_b^2-l^2r_b^2 + 3n^4-l^2n^2 + 4ql^2vn+q^2l^2)(l^2r_b ) by imposing condition ( b ) together with the regularity of the one - form @xmath30 one gets @xmath54 as the solution of the following fourth order algebraic equation 30nr_b^4 - 2r_b^3l^2+(6nl^2 + 27nl^2v^2 - 30n^3)r_b^2 - 27n^3l^2v^2=0 taking the limit @xmath74 one can obtain the asymptotically locally flat version of the above solutions which represents also new solutions of einstein - maxwell equations in six dimensions . here @xmath34 takes the following form f(r)= & & -the one - form potential @xmath30 and the function @xmath34 are the same as the ads case . for the nut case the mass parameter , @xmath35 , and the charge , @xmath1 , must acquire the following form m_n=-4 n^33q=-8n^3vimposing the regularity condition for @xmath30 and shift @xmath75 , one gets the condition for getting a nut solution \|v\| < for the taub - bolt case imposing the general conditions a ) and b ) together with the regularity condition on the potential @xmath30 the horizon radius satisfy the following algebraic equation 2r_b^3 - 3n(2 + 9v^2)r_b^2 - 27n^3v^2=0again the analytic solution of this equation is a relatively complicated expression so , instead , we plot @xmath54 in [ root2plot ] as a function of @xmath2 , where @xmath60 one can see clearly that @xmath76 . using @xmath63 as a base space , the metric of the charged taub - nut - ads solution has the form ds^2= f(r)(d+ 2nd_1 + 2nd_2)^2 + f(r)^-1dr^2+(r^2-n^2)(d_1 ^ 2+^2d_1 ^ 2+d_2 ^ 2+^2d_2 ^ 2 ) a = g(r)(d+2nd_1 + 2nd_2)where @xmath31 is given by g(r)= q + v here @xmath34 has the same form as the one for @xmath78 case . this six dimensional solution is a generalization to the solution found in @xcite . another exotic form similar to these metrics has been found in @xcite . also , these uncharged solutions have been used to find some bubble solutions in @xcite . for @xmath77 case taub - nut solution have curvature singularities at @xmath42 , indeed , r^ijklr_ijkl ~ ( r - n)^-2but the taub - bolt cases for such solution do not suffer from such a problem since the singularity is hidden behind the horizon . we have presented a class of @xmath0dimensions nut - charged solutions for einstein - maxwell field equations . these solutions depend on two extra parameters other than the mass and the nut charge , namely ; the charge @xmath1 and the potential at infinity @xmath2 . the existence of the parameter @xmath2 enables us to get a regularity condition on the one - form potential which we found identical to that required to obtain a nut solution . the nut conditions show that the mass not only depends on the nut charge as in four dimensions but also on the charge / or the potential @xmath2 as a result we get a smooth six dimensional manifold for @xmath79 . one might be concerned by the existence of naked singularities for the @xmath77 case . however , one would argue that from the point of view of the ads / cft duality that it is possible to resolve these singularities , for example , using gubser s @xcite criterion that these are ` good ' singularities , in the sense that a finite temperature deformation will yield a non - singular solution which is the bolt . it will be interesting to probe the physics of these new smooth gravitational solutions especially in the context of ads / cft correspondence and their thermodynamics . one can generalize these constructions easily to eight and ten dimensions . it is interesting also to generalization this construction to include black holes with topological horizons and study the nut and bolt conditions . + * acknowledgements * + i would like to the thank the abdus salam ictp for its hospitality and continues support , this is where this work was accomplished . i would like to thank edward teo and nidal chamoun for useful discussions , especially nc for suggestions and comments on a preliminary draft of this paper . 99999 a. h. taub , annal . math . * 53 * 471 ( 1951 ) ; e. newman , l. tamborino and t. unti , j. math . phys . * 4 * 915 ( 1963 ) . d. brill , phys . rev . * 133 * b845 . c. w. misner , j. math . phys . * 4 * 924 ( 1963 ) . d. j. gross and m. j. perry , nucl b226 * , 29 ( 1983);r . sorkin , phys.rev.lett.5187 ( 1983 ) . s.w . hawking , c.j . hunter and d. n. page , phys.rev.d59044033 ( 1999 ) , hep - th/9809035 . s.w . hawking and c.j . hunter , phys.rev . * d59 * 044025 ( 1999 ) , hep - th/9808085 . c.j . hunter , phys.rev.d59024009 ( 1999 ) , gr - qc/9807010 . c. v. johnson , _ d - brane primer _ , 1999 tasi summer school in strings , branes and gravity , hep - th/0007170 n. alonso - alberca , p. meessen and t. ortin , class.quant.grav . * 17 * 2783,hep - th/0003071 c. v. johnson and r. c. myers , phys.rev . * d50 * 6512 ( 1994 ) ; phys.rev . * d52 * 2294 ( 1995 ) . s. bais and p. batenberg , nucl . phys . * b253 * 162 ( 1985 ) d. n. page and c. n. pope , class . . grav . * 4 * 213 ( 1987 ) . a. chamblin , r. emparan , c. v. johnson , r. c. myers , phys.rev . * d59 * 064010 ( 1999),hep - th/9808177 . mann , phys.rev . * d60 * 104047 ( 1999),hep - th/9903229 . a. awad and a. chamblin , class . quant . grav . * 19 * 2051 ( 2002 ) . r. clarkson , l. fatibene and r.b . mann , nucl . * b652 * 348 ( 2003 ) , hep - th/0210280 ; r. mann , and c. stelea , hep - th/0408234 . a.m. ghezelbash and r.b . mann , jhep 0209:045 ( 2002 ) , hep - th/0207123 r. mann and cristian stelea , class.quant.grav . * 21 * 2937 ( 2004 ) , hep - th/0312285 . a. chamblin , r. emparan , c. v. johnson and r. c. myers , phys.rev . * d60 * 064018 ( 1999 ) , hep - th/9902170 . r. mann and cristina stelea , hep - th/0508186 . gubser , adv . theor . math . phys . * 4 * 679 ( 2002 ) , hep - th/0002160	we present a class of higher dimensional solutions to einstein - maxwell equations in @xmath0dimensions . these solutions are asymptotically locally flat , de sitter , or anti de sitter space - times . the solutions we obtained depend on two extra parameters other than the mass and the nut charge . these two parameters are the electric charge @xmath1 and the electric potential at infinity , @xmath2 , which has a non - trivial contribution . we analyze the conditions one can impose to obtain taub - nut or taub - bolt space - times , including the four - dimensional case . we found that in the nut case these conditions coincide with that coming from the regularity of the one - form potential at the horizon . furthermore , the mass parameter for the higher dimensional solutions depends on the nut charge and the electric charge or the potential at infinity .
a self - avoiding walk ( saw ) of length @xmath6 on a periodic graph or lattice @xmath7 is a sequence of distinct vertices @xmath8 in @xmath7 such that each vertex is a nearest neighbour of its predecessor . in figure [ fig : saw ] a very long walk of @xmath9 steps is shown ( generated by a monte carlo algorithm @xcite ) . steps on the square lattice courtesy of nathan clisby.,width=188 ] consider now those saws starting at the centre of an @xmath10 rectangle and ending when they hit the boundary . we take the scaling limit of saw , at the critical point , in the usual way ( described , for example , in @xcite ) . let @xmath11 be the probability that a randomly chosen saw , in the scaling limit , hits the side of the rectangle before the end , and let @xmath12 be the probability that it hits the end before the side . in guttmann and kennedy @xcite the probability ratio @xmath13 was calculated , under the assumption that the scaling limit is describable by @xmath14 for a rectangle of aspect ratio @xmath15 with parameter @xmath16 where @xmath17 they found , asymptotically , @xmath18,\ ] ] for @xmath19 where @xmath20 the exact result is given as the ratio of two integrals ( [ rat ] ) , and we give an outline of their calculation in the following section . in this work , we have considered all @xmath21 rectangles of aspect ratio 2 of size @xmath22 up to @xmath23 and have generated all saw starting at the origin and ending on the boundary of the rectangles . we have also considered all @xmath24 rectangles of aspect ratio 10 of size @xmath25 up to @xmath26 . we have also developed a refined enumeration scheme in which the full hitting distribution is recorded ; we give the results here for rectangles of aspect ratio 2 . the relevant geometry is shown in figure [ fig : lattice ] . we obtained separately the generating functions for saws starting at the centre of the rectangle and exiting at any prescribed point on its boundary . this has been compared to that calculated by conformal mapping , assuming that @xmath27 is the appropriate scaling limit . the data was generated using a transfer matrix formalism , utilising symmetry and similar techniques to minimise the computational complexity . details of the methods used are given in section [ series ] . in addition to the position of the end - points of the walks on the boundary , the length ( number of steps ) was also retained . in @xcite a monte carlo study was made by kennedy , who compared certain random variables associated with saws in a half - plane with those predicted assuming that the scaling limit of saws is given by @xmath14 excellent agreement was found in a visual comparison of two graphical plots . our work differs in nature in that we predict a specific value of @xmath2 , assuming that the scaling limit of saws is given by @xmath28 let @xmath29 be a bounded , simply connected domain in the complex plane containing @xmath30 we are interested in paths in @xmath29 starting at @xmath31 and ending on the boundary of the domain . initially we will consider random walks , later self - avoiding walks . we can discretize the space with a lattice of lattice spacing @xmath32 in both the random walk case and the saw case we are then interested in the scaling limit @xmath33 for random walks the scaling limit is brownian motion , stopping when it hits the boundary of @xmath34 the distribution of the end - point is harmonic measure . if the domain boundary is piecewise smooth , then harmonic measure is absolutely continuous with respect to arc length along the boundary @xcite . let @xmath35 denote the density with respect to arc length ( often called the poisson kernel ) . if @xmath36 is a conformal map on @xmath29 that fixes the origin , and such that the boundary of @xmath37 is also piecewise smooth , then the conformal invariance of brownian motion implies that the density for harmonic measure on the boundary of @xmath37 is related to the boundary of @xmath29 by @xmath38 in the case of the saw , lawler , schramm , and werner @xcite predicted that the corresponding density of the probability measure @xmath39 transforms under conformal maps as @xmath40 where @xmath41 and the constant @xmath42 is required to ensure that @xmath43 is a probability density . if one starts the random walk or the saw at the centre of a disc , then the hitting density on the circle will be uniform . so the above equations determine the hitting density for any simply connected domain , by virtue of the riemann mapping theorem . simulations in @xcite , @xcite and @xcite provide strong support for the conjectured behaviour . ( [ cisaw ] ) is correct for domains whose boundary consists only of vertical and horizontal line segments . for general domains there is a lattice effect that persists in the scaling limit that produces a factor that depends on the angle of the tangent to the boundary that must be included @xcite . ) the solution of the problem for random walks by conformal maps is described in @xcite , where a conformal map from the unit disc to an @xmath44 rectangle is given by the schwarz - christoffel formula . our approach also uses conformal maps , but we instead use a map between the upper half - plane and a rectangle , where the mapping is again given by a schwarz - christoffel transformation . for @xmath45 let @xmath46 @xmath47 is a schwarz - christoffel transformation that maps the upper half plane to a rectangle . the rectangle has one edge along the real axis and @xmath31 is a midpoint of this side . so the corners can be written as @xmath48 and @xmath49 where @xmath50 are the length of the horizontal and vertical edges , respectively . we have @xmath51 so @xmath52 we note that @xmath53 here @xmath54 is the complete elliptic integral of the first kind . by dilation invariance we only need concern ourselves with the aspect ratio @xmath55 so given an aspect ratio @xmath56 , we have to find @xmath57 such that @xmath58 from the properties of elliptic integrals , it follows @xcite that @xmath59 where @xmath60 is the jacobi theta function . evaluating this in one s favourite algebraic package gives the required value of @xmath57 for any @xmath61 instantly . alternatively , we note that for an aspect ratio of 10 , @xmath57 is very close to 1 . we can achieve very high accuracy by expanding the ratio of the above integrals around @xmath62 , and find @xmath63 solving this numerically for @xmath64 gives @xmath65 to leading order one obtains @xmath66 which for @xmath64 gives 19 digit accuracy . an alternative approach is to note that @xmath67 so that when @xmath68 one has @xmath69 here @xmath70 are _ singular moduli , _ that is , elliptic moduli for which @xmath71 from a table of singular moduli , for example @xcite , one finds for @xmath64 that @xmath72 for aspect ratio @xmath73 one finds @xmath74 these considerations establish @xmath75 a parameter of the conformal map , both directly and asymptotically . to proceed further we need to determine the behaviour of the hitting density of paths from the origin to the boundary of the rectangle . first , note that the preimage of the center of the rectangle will , by symmetry , be on the imaginary axis , so write it as @xmath76 in @xcite geometrical arguments are given to show that @xmath77 an alternative algebraic derivation is the following : from the schwarz - christoffel mapping @xmath78 set @xmath79 giving @xmath80 this integral is an ( incomplete ) elliptic integral of the second kind , from which follows @xmath81 as shown in @xcite , for both the random walk and the saw in the half plane starting at @xmath82 the ( unnormalized ) hitting density along the real axis is @xmath83 . hence the hitting density @xmath84 for walks in a rectangle starting at the center is @xmath85 the ratio of probabilities of a walk first hitting an end of the rectangle to that of a walk first hitting a side is just the ratio of the integral of the hitting density @xmath39 along a vertical edge to the integral along a horizontal edge , @xmath86 by a change of variable , setting @xmath87 in the denominator and @xmath88 in the numerator , and recalling that @xmath89 the ratio of probabilities of a first hit on the vertical side to a first hit on the horizontal side , @xmath90 , is @xmath91 these integrals can not be solved in closed form . both mathematica and maple can be used to evaluate them numerically to any reasonable accuracy . however for our purposes it would be useful to have a reasonably accurate asymptotic representation . we will calculate the leading order and first correction term in the quotient ( [ rat ] ) where @xmath92 $ ] and @xmath93 with @xmath94 here @xmath57 is related to the aspect ratio @xmath56 of a rectangle by ( [ alfa ] ) . consider the numerator first . one has @xmath95 now , since @xmath96 with @xmath97 @xmath98 so @xmath99 @xmath100 set @xmath101 then @xmath102 and @xmath103^{\frac{b-1}{2 } } dt - \frac{bc}{2 } \cdot \int_0 ^ 1 ( 1 + 2t)[t(1-t)]^{\frac{b-1}{2 } } dt\right ) \nonumber \\ = & \left ( 1+\frac{c(b-1)}{4 } \right ) \left ( ( \alpha-1)^b \frac{\sqrt{\pi } \cdot \gamma \left ( \frac{1}{2}+\frac{b}{2 } \right ) } { 2^{b+1 } \cdot \gamma \left ( 1+\frac{b}{2 } \right ) } ( 1+\frac{b-1}{4}c)- ( \alpha-1)^{b+1 } \frac{b \cdot \sqrt{\pi } \cdot \gamma \left ( \frac{1}{2}+\frac{b}{2 } \right ) } { 2^{b+1 } \cdot \gamma \left ( 1+\frac{b}{2 } \right ) } \right ) \nonumber \\ = & ( \alpha-1)^b \frac{\sqrt{\pi } \cdot \gamma \left ( \frac{1}{2}+\frac{b}{2 } \right ) } { 2^{b+1 } \cdot \gamma \left ( 1+\frac{b}{2 } \right ) } \left [ 1 - \frac{(b+1)}{2}(\alpha-1)\right ] . \end{aligned}\ ] ] next , consider the denominator : @xmath104^{(b-1)/2 } du\ ] ] again , let @xmath93 where @xmath94 one has @xmath105 and @xmath106 both to first order in @xmath107 so with @xmath108 @xmath109 where @xmath110^{(b-1)/2 } dx.\ ] ] ( note that the first argument refers to the power of @xmath111 only , not the other occurrences of @xmath112 ) . next , we take the mellin transform @xmath113 so the mellin transform of @xmath114 is @xmath115 @xmath116 hence the inverse transform is @xmath117 where @xmath118 closing the contour to the left , as we want a small @xmath119 expansion , we see there are poles coming from the gamma functions . the first occurs at @xmath120 and the second at @xmath121 taking the residue at the poles we get @xmath122 another term comes from @xmath123 and this gives a term @xmath124 we can eliminate the hypergeometric functions , as they can be expressed in terms of gamma functions , and then simplified . in this way we obtain @xmath125.\ ] ] we can write the ratio @xmath126 = \ ] ] @xmath127 here we have used the asymptotic expansion ( [ alfa ] ) of @xmath128 as an expansion in @xmath75 and the parameters introduced above are @xmath129 and @xmath130 @xmath131 @xmath132 then @xmath133,\ ] ] for @xmath134 for @xmath64 and @xmath135 the leading term gives 4 significant digits accuracy ( where we are comparing to the value obtained by numerical integration with 50 digit accuracy ) , while including the next terms gives 8 significant digits . this is more than sufficient for our purpose here . however for @xmath136 ( with the same value of @xmath112 ) the expansion is much more slowly convergent , and the above expression provides the result to an accuracy of merely 1 part in 20 . while we could calculate further terms in the asymptotic expansion , we prefer to simply evaluate the integral expression ( [ rat ] ) numerically . we now use these results to calculate the value of @xmath112 characterising the scaling limit of saws . we do this by generating all saws in a rectangle of given aspect ratio @xmath56 , centred at the origin , and of size @xmath137 for @xmath138 where @xmath139 refers to the largest possible rectangle we can use given our computer resources . in the next section we describe the efficient generation of these saws , using transfer matrices and efficient use of symmetry and the elimination of unnecessary starting points . in the following section we analyse the data thus produced , by calculating and extrapolating the ratio @xmath140 for two representative values of @xmath141 which were @xmath136 and @xmath142 from these estimates of @xmath143 we use the asymptotic expansion above to estimate @xmath16 in the case of aspect ration 10 , while for @xmath136 the comparison is made with numerical integration values from ( [ rat ] ) . furthermore , we can compare the hitting density distribution function , as calculated from our largest rectangle , to that predicted in the scaling limit . with @xmath144 and for rectangles of aspect ratio 2 , it will be seen that the two distribution functions are almost indistinguishable . we first describe the computation of the generating functions of a saw starting at the centre of the strip and terminating , respectively , the first time it hits either one of the rectangle s two short sides , or one of its two long sides . the rectangle is supposed oriented so that the short sides are horizontal and the long sides vertical . as usual in enumeration studies the basic ingredient is a transfer matrix acting on an appropriate set of states . each state comes with a weight , which is a polynomial with integer coefficients in a variable @xmath145 that is interpreted as the fugacity for one step of the saw . the transfer matrix is maximally decomposed as a product of elementary matrices that each adds a small piece ( a vertex and its adjacent half - edges ) to the lattice being constructed . the states contain information about how the uppermost row of half - edges are connected among themselves , or to the boundary of the rectangle , or to the centre point , through pieces of the partially built saw . this information and the action of the elementary matrices on the states must be carefully devised so that : 1 . a single saw going from the centre to the boundary of the rectangle is constructed ; 2 . the two desired generating functions , describing respectively the hitting of the short and long sides of the rectangle , can be disentangled ; 3 . no superfluous information is stored . points 1 and 2 are essential for the correctness of the algorithm , whereas 3 is only a matter of efficiency . it can not be excluded that the precise procedure described below can be shown in the future not to be optimal , in the sense that a variant algorithm can be constructed that uses fewer states to achieve the goals 1 and 2 . we consider an @xmath10 rectangle , with @xmath146 and @xmath147 even ( so that the centre of the rectangle coincides with a vertex ) , and @xmath148 ( the transfer matrix then propagates upwards ) . the states contain information about the connectivities of the @xmath6 uppermost half - edges , labelled @xmath149 from left to right , of the partially constructed lattice . when a row of the lattice is completed , @xmath150 and all half - edges are vertical . at stages where a row is under construction , @xmath151 and one of the half - edges is horizontal . the lattice is constructed through the action of the elementary matrices row by row , completing the rows from left to right . this is shown in figure [ fig : tm ] . each half - edge can be connected to another half - edge , or be in one of the following configurations : 1 . empty , meaning that it is not covered by the saw ; 2 . side , when it is connected through the saw to the left or right boundary of the rectangle ; 3 . bottom , when its is connected to the bottom ( or top ) boundary of the rectangle ; 4 . centre , when it is connected to the centre point . in addition the system as a whole can be in the following configurations when the saw has been completed before the lattice has been completely built : 1 . exit_lr when the saw went from the centre point to the left or right boundary ; 2 . exit_bt when it went to the bottom or top boundary . being in one of those exit configurations implies that all the half - edges must be empty and remain so until completion of the lattice . to describe such states it is therefore sufficient to use just @xmath152 label containing the exit status ( exit_lr or exit_bt ) . imposing the correct constraints throughout the construction of the lattice ( see below ) will imply that once the lattice has been completed only the two exit states carry a non - zero weight . these two weights are precisely the desired generating functions . ( 0,0 ) grid ( 4,6 ) ; ( 0,0 ) ( 4,0 ) ( 0,6 ) cycle ; ( 2,3 ) circle ( 1ex ) ; in 1,2,3 ( , 0 ) ( , 0.5 ) ; ( 2,0 ) node[below]start ; ( 0,0 ) grid ( 4,6 ) ; ( 0,0 ) ( 4,0 ) ( 4,6 ) ( 0,6 ) cycle ; ( 2,3 ) circle ( 1ex ) ; in 1,2,3 ( , 0 ) ( , 1.5 ) ; ( 0,1 ) ( 4,1 ) ; ( 0,2 ) ( 0.5,2 ) ; ( 2,0 ) node[below]left ; ( 0,0 ) grid ( 4,6 ) ; ( 0,0 ) ( 4,0 ) ( 4,6 ) ( 0,6 ) cycle ; ( 2,3 ) circle ( 1ex ) ; in 1,2,3 ( , 0 ) ( , 1.5 ) ; ( 0,1 ) ( 4,1 ) ; ( 0,2 ) ( 1.5,2 ) ; ( 1,1.5 ) ( 1,2.5 ) ; ( 2,1.5 ) ( 2,2.5 ) ; ( 1.5,2 ) ( 2.5,2 ) ; ( 2,0 ) node[below]bulk ; ( 0,0 ) grid ( 4,6 ) ; ( 0,0 ) ( 4,0 ) ( 0,6 ) cycle ; ( 2,3 ) circle ( 1ex ) ; in 1,2,3 ( , 0 ) ( , 2.5 ) ; ( 0,1 ) ( 4,1 ) ; ( 0,2 ) ( 3.5,2 ) ; ( 3.5,2 ) ( 4,2 ) ; ( 2,0 ) node[below]right ; ( 0,0 ) grid ( 4,6 ) ; ( 0,0 ) ( 4,0 ) ( 0,6 ) cycle ; ( 2,3 ) circle ( 1ex ) ; in 1,2,3 ( , 0 ) ( , 2.5 ) ; in 1,2 ( 0 , ) ( 4 , ) ; ( 0,3 ) ( 1.5,3 ) ; ( 1,2.5 ) ( 1,3.5 ) ; ( 2,2.5 ) ( 2,3.5 ) ; ( 1.5,3 ) ( 2.5,3 ) ; ( 2,3 ) circle ( 1ex ) ; ( 2,0 ) node[below]centre ; ( 0,0 ) grid ( 4,6 ) ; ( 0,0 ) ( 4,0 ) ( 4,6 ) ( 0,6 ) cycle ; ( 2,3 ) circle ( 1ex ) ; in 1,2,3 ( , 0 ) ( , 5.5 ) ; in 1,2,3,4,5 ( 0 , ) ( 4 , ) ; ( 1,5.5 ) ( 1,6 ) ; ( 2,3 ) circle ( 1ex ) ; ( 2,0 ) node[below]top ; we now describe the action on the states of the different elementary matrices that build up the system ( see figure [ fig : tm ] ) . in the initial stage , called start in figure [ fig : tm ] , the system is in a superposition of the completely empty state ( with weight @xmath153 ) , and each of the @xmath6 states where one of the @xmath150 half - edges carries the @xmath154 label ( with weight @xmath145 ) and the others are empty . note that a valid state can not have more than one half - edge labelled either bottom or side , since we aim at constructing a saw that touches the boundary in only one single point . after a row of the lattice has been completed the following operation , called left in figure [ fig : tm ] , adds a horizontal half - edge touching the left boundary of the rectangle . if the system is already in an exit state it remains so and with the same weight . otherwise the labels are shifted one unit to the right to make place for that of the horizontal half - edge . the latter can either be empty or , if no other bottom or side label exists , be set to side in which case the weight is multiplied by @xmath145 to take account for the extra monomer . the subsequent operations within the row are of the type bulk . the elementary matrix now acts on two labels , describing the half - edges coming from west and south in the incoming state , and which after the time evolution describe the half - edges going towards north and east in the outgoing state . an exit state is treated as before . in all other cases we must ensure that the bulk vertex is adjacent to either two or four occupied edges . thus , if both incoming labels are empty , the outgoing labels can be empty as well , or mutually connected with an extra weight @xmath155 . if only one incoming label is empty , the other can go straight through the vertex or make a turn , in both cases with weight @xmath145 . finally , if neither of the incoming labels is empty there is a rather large number of possibilities : 1 . if both labels describe connectivities to other half - edges ( i.e. , neither is equal to one of the special labels empty , side , bottom or centre described above ) , their respective partner half - edges must be mutually connected . however , if the connection is among the two incoming half - edges themselves , the state must be discarded , since the saw is not allowed to form a closed loop . if one of the labels is special , the partner of the non - special label becomes equal to the special label . 3 . finally , if both incoming labels are special , we have the opportunity of creating an exit state . but first a number of precautions must be taken : 1 . all the remaining labels in the state need to be empty , otherwise the state must be discarded . one of the incoming labels must be centre and the other either side or bottom ( otherwise discard the state ) . the latter label then determines whether the exit state is exit_lr or exit_bt . in all those cases the two outgoing labels will be empty and the weight coincides with that of the incoming state . to complete a row we use the operation right . the elementary matrix acts on the rightmost label , describing the half - edge that will touch the right boundary of the rectangle . the action on exit states is as usual . if the incoming label is empty the state is kept as it is . otherwise the rightmost half - edge will be occupied , and we must first make sure that there is not already a label which is side or bottom ( otherwise discard the state ) . next , if the incoming label is centre we obtain an exit_lr exit state provided that the remainder of the half - edges are empty . otherwise , if the incoming half - edge is connected to another half - edge , the latter will be marked by a side label to signify the hitting of the right boundary of the rectangle . after these considerations the rightmost label is deleted , since we have now completed a row of the lattice and only @xmath156 labels are required . special care must be taken when attaining the centre point . this is achieved by applying the operation centre . as in the bulk case the action is on two labels , but while in the former case the vertex had degree two or four , the number of occupied adjacent edges must now be one , since the saw emanates from the centre point . thus , if both incoming labels are empty we produce two outgoing states with one of the outgoing labels being centre and the other empty ( with a weight @xmath145 ) . states with two occupied ingoing labels must be discarded . and finally , for states with one ingoing label being empty and the other occupied there are a number of possibilities : 1 . if the occupied incoming half - edge is connected to another half - edge , the latter acquires the centre label . otherwise the occupied incoming half - edge is of the type side or bottom , and we end up in an exit state ( provided the remainder of the system is empty ) . it remains to describe how to finish the rectangle after building the last row . each half - edge must now be subjected to the operation top in order to implement the top boundary of the rectangle . the corresponding elementary matrix acts on a single label ( and is in fact very similar to the operation right described above ) . if that label is empty the state is unchanged . if it is side or bottom the state is discarded . if it is connected to another half - edge , the latter acquires the bottom label ( which is ill - named in this case , since it is the top boundary of the rectangle which is being touched ) . and finally , if the incoming label is centre we end up in the exit_bt state ( provided the remainder of the system is empty ) . from a practical point of view , all the labels used to describe a state can be coded on short integers ( say , of the char type in c ) and the coding of the entire state of @xmath6 half - edges is then an array of characters . the handling of incoming and outgoing states is then conveniently and efficiently done by employing standard hashing techniques . the integer coefficients entering the weights will obviously become very large for even a moderately - sized system , but this is easily dealt with by using modular arithmetic , i.e. , repeating the entire computation modulo various primes and reconstituting the results from the chinese remainder theorem . our results for the generating functions with various aspect ratios are so lengthy that it makes little sense to have all them all appear in print . however , the complete results are available in electronic form as supplementary material to this paper . for the convenience of the readers who wish to check explicitly the example of the @xmath157 rectangle shown in figure [ fig : tm ] we give the results for that case . for the hitting of the long sides of the rectangle we find @xmath158 while for the hitting of the short sides we have @xmath159 note that all the coefficients in this example are even . this is generally true and due to the symmetries of the rectangle . explicitly , the mirror symmetry with respect to a horizontal ( resp . vertical ) line going through the centre point induces a bijection between the saws that hit the bottom ( resp . left ) boundary and those that hit the top ( resp . right ) boundary . we can take advantage of those symmetries to make a more efficient version of the algorithm . namely , if we impose the restriction that all the half - edges incident on the bottom and left boundaries be empty , only the saws hitting the right and top boundaries will be counted . this means that the required generating functions are simply divided by a factor of two . the reason that we have imposed the constraint on the left boundary rather than the right , and on the bottom boundary rather than the top , is that the left and bottom boundaries are those encountered earliest in the time evolution of figure [ fig : tm ] . imposing the constraints early on will most efficiently curb the proliferation of unnecessary states . .maximum number of states in the transfer matrix algorithm for an @xmath10 rectangle with @xmath160 . here @xmath161 ( resp . @xmath162 ) refers to the time evolution before ( resp . after ) the addition of the centre point . the corresponding primed quantities , @xmath163 and @xmath164 , are for the improved algorithm that exploits reflection symmetries . [ cols=">,>,>,>,>",options="header " , ] the need to register the exit coordinate obviously increases the number of states used in the computation . one might expect that roughly @xmath165 times more states would be needed . fortunately , the additional constraints imposed on the boundary half - edges ameliorates the situation somewhat . the maximum number of states needed by the full hitting statistics algorithm are shown in table [ tab : states - full ] . in table [ table : aspect2 ] we give the ratio @xmath166 of the number of walks hitting the long side to that of the number of walks hitting the short side for a rectangle with aspect ratio 2 . c c @xmath6 & ratio @xmath166 + [ 0.5ex ] 2 & 4.6381585303417408684303075667444130488805022010318359737078706077696 + 4 & 4.5626403997998714832892051885980313566654043362375413463572606810903 + 6 & 4.5737694425659079263885691980864259391038701844156223313634364418527 + 8 & 4.5816876116611105157124235041836861179217970948263182063453368731855 + 10 & 4.586835597801674598789736657813550866717170851432063193366000200424 + 12 & 4.590372957831729906013721282115051244422423169272971142266883828004 + 14 & 4.592938409379941423415363287957199443993269739941033366762062737428 + 16 & 4.594881296788588474819955070086767540896832163815763639405593859180 + 18 & 4.5964037339238675392443305096632781995673848727512239850106566816295 + [ 1ex ] [ table : aspect2 ] this has been calculated by evaluating the ratio of the two generating functions " ( which are of course polynomials , as we are dealing with finite lattices ) at @xmath167 for the square lattice @xmath168 is not exactly known , but the best numerical estimate , @xmath169 @xcite , is in agreement with the mnemonic @xmath170 @xcite , and it is the appropriate root of this polynomial equation that we are using . our results are insensitive to the uncertainty in the best numerical estimate of @xmath167 this ratio is seen to be monotonically decreasing , and a plot of the data suggests behaviour of the form @xmath171 in order to estimate @xmath172 we notice that a log - log plot of @xmath6 against @xmath173 has gradient @xmath174 this implies @xmath175 and in our subsequent analysis we assume @xmath176 we then extrapolated the sequence @xmath177 by the bulirsch - stoer algorithm @xcite , with the free parameter in that algorithm set to reflect @xmath178 the analysis is shown in table [ table : extrap ] . c c c c c c c 4.589060 & 4.694672&4.609872 & 4.608214 & 4.608850 & 4.609984 & 4.609630 + 4.600917&4.615954&4.608989 & 4.608787 & 4.608949 & 4.609718 & + 4.604615&4.611393 & 4.608899 & 4.609077 & 4.608726 & & + 4.606178&4.610139 & 4.608975 & 4.609228 & & & + 4.606991&4.609677 & 4.609086 & & & & + 4.607480&4.609489 & & & & & + 4.607806 & & & & & & + [ 1ex ] [ table : extrap ] in this way we estimate the limiting ratio as @xmath179 we also analysed the data by several other appropriate extrapolation algorithms . the brezinski @xmath180 algorithm @xcite gives an increasing sequence of estimates for the large @xmath6 limit of the ratios , sugesting that the limit @xmath181 levin s @xmath182-transform @xcite gives a monotone decreasing sequence of estimates , suggesting that @xmath183 neville tables @xcite also give a monotone increasing sequence of estimates , suggesting that @xmath184 these are all consistent with our estimate from the bulirsch - stoer algorithm . this gives @xmath185 or @xmath186 in table [ table : aspect10 ] we give the corresponding , slightly shorter data set , for walks in a rectangle of aspect ratio 10 . by a similar analysis , we estimate the limit to be @xmath187 from ( [ asympt ] ) this implies @xmath188 or @xmath189 c c @xmath6 & ratio @xmath190 + [ 0.5ex ] 4 & 14006.18549331435655361766127203880086492399593549150863381964074817 + 6 & 14245.30058306730412888413593536062491411218703615652940541335814143 + 8 & 14391.29165062927743781254092745976155303961361728983232311816771246 + 10 & 14487.00740644606426682399787816797507371639547391783093463818418043 + 12 & 14554.35495421800301499956345001503940598563417509710466751030105068 + 14 & 14604.3566053623407864953033765640802317538963314516614175885263116878 + [ 1ex ] [ table : aspect10 ] as well as calculating the ratio of the number of walks hitting the long side to that of the number of walks hitting the short side of a rectangle with aspect ratio 2 , we have also calculated the distribution of hitting densities along the long side of such a rectangle . for the case of a @xmath191 rectangle , the data is shown in table [ table : hitden ] . from ( [ cisaw ] ) it follows that the ( unnormalized ) probability density satisfies @xmath192 where @xmath193 follows from the schwarz - christoffel transformation previously discussed , and the unrenormalized hitting density along the real axis is @xmath194 as follows from ( [ cirw])([cisaw ] ) , and is described in detail in @xcite . so the hitting density along the long side of the rectangle is ( up to a multiplicative constant ) , @xmath195 where the corresponding ordinates follow from the schwarz - christoffel transformation , so that the ordinates @xmath196 transform as @xmath197 where @xmath198 is the incomplete elliptic integral of the first kind . in figure [ fig : hit - dens ] we show the predicted hitting density data for a rectangle of aspect ratio 2 together with the measured data for saws in a @xmath191 rectangle . the enumeration data has been scaled linearly to make the origin and end - points coincide . despite the fact that a @xmath191 rectangle would seem to be rather far from the scaling limit , the agreement is remarkably good . c c c @xmath6 & @xmath199 & hitting number + [ 0.5ex ] 0 & 0.000000000000000000000000000000 & 0.372179965985066951892060536379 + 1 & 0.071428571428571428571428571428 & 0.365592202518487166323769503648 + 2 & 0.142857142857142857142857142857 & 0.347584595139263204799699150831 + 3 & 0.214285714285714285714285714286 & 0.321757459504506388412145573956 + 4 & 0.285714285714285714285714285714 & 0.291774774902074605895271622535 + 5 & 0.357142857142857142857142857143 & 0.260478252386315460357350977416 + 6 & 0.428571428571428571428571428571 & 0.229709762715325246279406382177 + 7 & 0.500000000000000000000000000000 & 0.200461104152704918984187699069 + 8 & 0.571428571428571428571428571429 & 0.173099730010024355136783990185 + 9 & 0.642857142857142857142857142857 & 0.147556785152232511539056308909 + 10 & 0.714285714285714285714285714286 & 0.123439559572782004646612438472 + 11 & 0.785714285714285714285714285714 & 0.100052601022725874927123950740 + 12 & 0.857142857142857142857142857143 & 0.076292442893068500584108376384 + 13 & 0.928571428571428571428571428571 & 0.050321628007048540057861695734 + [ 1ex ] [ table : hitden ] rectangle ( green ) compared to the theoretical prediction in the scaling limit with @xmath135 ( red).,width=302 ] we first calculated the ratio of the number of saws starting at the centre of a rectangle and hitting the end to that of the number of saws starting at the centre of a rectangle and hitting the sides , for rectangles of a given aspect ratio . then by comparing this to the calculated value @xcite assuming that the scaling limit of saws is given by @xmath200 we estimated @xmath2 for rectangles of aspect ratio 2 and 10 . in the former case we found @xmath201 and in the latter case we found @xmath202 this would appear to be the strongest numerical support yet for the widely held belief that the scaling limit of saws is given by @xmath203 furthermore , we have also calculated the distribution of hitting densities along the long side of such a rectangle and compared this with the actual distribution for a @xmath191 rectangle . visually , the agreement between the plots of the two distributions is seen to be quite persuasive . this work was supported by the australian research council through grant dp120100939 ( ajg and jlj ) , and by the institut universitaire de france and agence nationale de la recherche through grant anr-10-blan-0414 ( jlj ) . ajg would like to thank larry glasser for a lesson in asymptotic expansions , tom kennedy for a lesson in conformal mapping , and both authors would like to gratefully acknowledge the support of the simons center for geometry and physics at suny stony brook where this work was , in part , carried out . lawler g , schramm o and werner w ( 2004 ) on the scaling limit of planar self - avoiding walk , _ fractal geometry and applications : a jubilee of benoit mandelbrot , part 2 _ , 339 - 364 , proc . pure math . 72 , amer . soc . , providence , ri ( arxiv : math/0204277v2 )	it is widely believed that the scaling limit of self - avoiding walks ( saws ) at the critical temperature is ( i ) conformally invariant , and ( ii ) describable by schramm - loewner evolution ( sle ) with parameter @xmath0 we consider saws in a rectangle , which originate at its centre and end when they reach the boundary . we assume that the scaling limit of saws is describable by @xmath1 with the value of @xmath2 to be determined . it has previously been shown by guttmann and kennedy @xcite that , in the scaling limit , the ratio of the probability that a saw hits the side of the rectangle to the probability that it hits the end of the rectangle , depends on @xmath3 by considering rectangles of fixed aspect ratio 2 , and also rectangles of aspect ratio 10 , we calculate the probabilities exactly for larger and larger rectangles . by extrapolating this data to infinite rectangle size , we obtain the estimate @xmath4 for rectangles of aspect ratio 2 and @xmath5 for rectangles of aspect ratio 10 . we also provide numerical evidence supporting the conjectured distribution of saws striking the boundary at various points in the case of rectangles with aspect ratio 2 . .3 cm
recent advances in micro - electromechanical systems ( mems ) and wireless communications have enabled the development of very small , smart , low cost sensing devices ( @xcite ) with sensing , data - processing and wireless transmission capabilities . they are meant to be pervasively deployed into forming ad - hoc wireless sensor networks that collect information from the ambient environment and make it available to the user . some applications imply deployment in remote or hostile environments ( battle - field , tsunami , earth - quake , isolated wild - life island , space exploration ) to assist in tasks such as target tracking , enemy intrusion detection , forest fire detection , environmental or biological monitoring . some other applications imply deployment indoors or in urban or controlled environments . examples of such applications are industrial supervising , indoor micro - climate monitoring ( e.g. to reduce heating cost by detecting poor building thermal insulation ) , smart - home applications , patient - doctor health monitoring or blind and impaired assisting . sensor networks imply distributed and collaborative data - processing , because of the small utility of each sensor individually and the severe resource constraints , mainly with respect to energy but also memory and computation capabilities . we wish to solve the problem of tracking objects moving in a domain over a certain period of time i.e. we want the wireless sensor network to be able to detect the position of any moving object in the domain at any time in the monitoring period . location tracking is the current standard approach to the tracking problem in sensor networks . location tracking includes several phases : target detection , distance estimation , position evaluation and trajectory estimation . the target detection is performed individually by sensor nodes in the network and does not need any synchronization . the distance estimation uses either an empiric law ( e.g. based on signal attenuation estimations ) translating the sensing intensity into a real valued distance , or uses a special hardware device performing this translation . the position estimation phase is typically addressed with trilateration . at least three nodes are periodically chosen among the sensors located next to the target . each of these three nodes gives an estimate of its distance to the target . finally , another node ( that can be one of the three ) performs trilateration based on the distance estimates and computes the position estimate . the position evaluation is then sent to a base - station in order to estimate the trajectory of the moving object . while this method can give very precise results when the network is dense enough ( at least 3 nodes have to be located next to the target as it moves ) , it has several drawbacks . first , this technique does not scale to the case of tracking multiple objects , for instance a group of objects moving in formation , or several individual objects . protocols designed to compute in a precise manner the position of an individual target are not able to cope with the question `` detect any of the members of a given group '' . second , the protocol overhead is high in terms of communication . each time a position is estimated , three nodes are chosen . then , among these three nodes another node is chosen to perform the position evaluation . this is done through exchange of protocol messages and can be quite energy consuming . while precision may be needed in some applications , other applications need only rough information on the location area . for instance , if the target is represented by a group of several objects , we may want to detect just an arbitrary location within this area . finally , this technique assumes the existence of a base station with powerful computational resources , generally supposed to be fixed . although this is sometimes realistic , many of the tracking applications do not respect this hypothesis . for instance , soldiers tracking an enemy would be moving to follow it . another example is habitat monitoring : a group of biologists tracking an antelope herd also needs to move towards the detected herd . in fact , many natural tracking applications actually use a mobile sink . in order to cope with these problems we broaden the hypothesis defining the tracking problem : a ) the target can be an individual object ( e.g. a patient in a hospital ) as well as a large group of several objects ( e.g. an animal herd ) . b ) the tracking precision becomes a parameter of the problem . we are primarily concerning ourselves with fuzzy tracking applicable to large group of objects , where the degree of precision needs not be high . but our techniques can use any sort of location information attached to sensors to enhance our solution s quality ( for instance , if in a hospital sensor nodes are aware of their location in terms of building wing and room number , and we can use this information for the tracking results ) . c ) the sink can be either fixed or mobile and there can be more than one sink ( for instance , troopers or a group of biologists ) . as the fixed sink case is rather well studied , we are concerning ourselves mainly with the mobile sink problem , which is useful in applications but also more challenging to cope with . the main idea of our approach is to avoid actively involving sensors in the tracking process . instead , we exploit `` traces '' of target presence that are anyway left around its moving trajectory . the role of sensors in our protocol is rather passive : they just locally decrease trace intensities with time ( to take into account the fact that the target was detected but then moved away ) and also propagate them appropriately , in order to spread then in a balanced way in the network at a low energy cost . the active role in our tracking approach is performed by the tracking agent , that greedily follows trace gradients to locate the moving target . as shown by the simulation findings , our approach achieves significant improvements over well known methods in the state of the art . first , our protocol is successful ( i.e. the target is indeed located ) , even in the case of multiple targets and mobile sinks . also , our protocol is very efficient , since it reduces energy a lot ( by avoiding active sensor participation ) while keeping latency low ( by spreading trace intensities in a way that `` covers '' the network in a balanced way ) . a standard centralized approach to tracking ( @xcite ) , is `` sensor specific '' , in the sense that it uses some smart powerful sensors that have high processing abilities . in particular , this algorithm assumes that each node is aware of its absolute location ( e.g. via a gps ) or of a relative location . the sensors must be capable of estimating the distance of the target from the sensor readings . the process of tracking a target has three distinct steps : detecting the presence of the target , determining the direction of motion of the target and alerting appropriate nodes in the network . thus , in their approach a very large part of the network is actively involved in the tracking process , a fact that may lead to increased energy dissipation . also , in contrast to our method that can simultaneously handle multiple targets , their protocol can only track one target in the network at any time . overall , their method has several strengths ( reasonable estimation error , precise location of the tracked source , real time target tracking , but there are weaknesses as well ( intensive computations , intensive radio transmissions ) . our method is entirely different to the network architecture design approach for centralized placement / distributed tracking ( see e.g. the book @xcite for a nice overview ) . according to that approach , optimal ( or as efficient as possible ) sensor deployment strategies are proposed to ensure maximum sensing coverage with minimal number of sensors , as well as power conservation in sensor networks . in one of the centralized methods ( @xcite ) , that focuses on deployment optimization , a grid manner discretization of the space is performed . their method tries to find the gridpoint closest to the target , instead of finding the exact coordinates of the target . in such a setting , an optimized placement of sensors will guarantee that every gridpoint in the area is covered by a unique subset of sensors . another network design approach for tracking is provided in @xcite , that tries to avoid an expensive massive deployment of sensors , taking advantage of possible coverage ovelaps over space and time , by introducing a novel combinatorial model ( using set covers ) that captures such overlaps . the authors then use this model to design and analyze an efficient approximate method for sensor placement and operation , that with high probability and in polynomial expected time achieves a @xmath0 approximation ratio to the optimal solution . clearly , in contrast to our direct approach to tracking , such network design solutions can provide full tracking only when combined with collaborative processing methods , to process and synthesize individual target location estimations . as opposed to centralized processing , in a distributed model sensor networks distribute the computation among sensor nodes . each sensor unit acquires local , partial , and relatively coarse information from its environment . the network then collaboratively determines a fairly precise estimate based on its coverage and multiplicity of sensing modalities . several such distributed approaches have been proposed . in @xcite , a cluster - based distributed tracking scheme is provided . the sensor network is logically partitioned into local collaborative groups . each group is responsible for providing information on a target and tracking it . sensors that can jointly provide the most accurate information on a target ( in this case , those that are nearest to the target ) form a group . as the target moves , the local region must move with it ; hence groups are dynamic with nodes dropping out and others joining in . it is clear that time synchronization is a major prerequisite for this approach to work . furthermore , this algorithm works well for merging multiple tracks corresponding to the same target . however , if two targets come very close to each other , then the mechanism described will be unable to distinguish between them . another nice distributed approach is the dynamic convoy tree - based collaboration ( dctc ) framework that has been proposed in @xcite . the convoy tree includes sensor nodes around the detected target , and the tree progressively adapts itself to add more nodes and prune some nodes as the target moves . in particular , as the target moves , some nodes lying upstream of the moving path will drift farther away from the target and will be pruned from the convoy tree . on the other hand , some free nodes lying on the projected moving path will soon need to join the collaborative tracking . as the tree further adapts itself according to the movement of the target , the root will be too far away from the target , which introduces the need to relocate a new root and reconfigure the convoy tree accordingly . if the moving target s trail is known a priori and each node has knowledge about the global network topology , it is possible for the tracking nodes to agree on an optimal convoy tree structure ; these are at the same time the main weaknesses of the protocol , since in many real scenarios such assumptions are unrealistic . finally , a `` mobile '' agent approach is followed in @xcite , i.e. a master agent is traveling through the network , and two slave agents are assigned the task to participate to the trilateration . as opposed to our method , their approach is quite complicated , including several sub - protocols ( e.g. election protocols , trilateration , fusion and delivery of tracking results , maintaining a tracking history ) . although by using mobile agents , the sensing , computing and communication overheads can be greatly reduced , their approach is not scalable in randomly scattered networks and also for well connected irregular networks , since a big amount of offline computation is needed finally , the base that receives the tracking results is assumed fixed ( in a tracking application this can be a problem ) . the interested reader is referred to @xcite , the nice book by f. zhao and l. guibas , that even presents the tracking problem as a `` canonical '' problem for wireless sensor networks . also , several tracking approaches are presented in @xcite . the intuitive idea behind our tracking protocol is inspired by a natural model : the way a tiger tracks an antelope in a savannah . antelope traces are initially stored into the environment as the antelope moves . the intensity of these initial traces decreases with time . these traces are partially spread through the savannah ( e.g. due to the wind action ) and the intensity of a spread trace decreases with the distance from the initial trace . thus , the only role of the environment is to store and to rather passively spread traces . it represents a passive actor of the tracking process . the active actor of the tracking process is the tiger , which senses the traces stored into the environment and tries to track the antelope by following the trace gradient . the correspondence to our problem is the following : the sensor network is the environment , the tiger is a tracking agent and the antelope is the tracking object . using this natural model has the advantage that as long as no tracking demand is issued , no ( or few resources ) are used into the network , as opposed to the location tracking algorithms where the network is continuously tracking the object in a proactive manner , hence consuming valuable energy . a passive network is the key to our energy optimisation . on the other hand , when the tracking is ( reactively ) in progress , only local computations and very light computation resources are needed in order to follow the trace gradient . when the tracking is needed , a tracking agent starts to walk through the network . this agent follows the trace gradient in a greedy manner . the agent can be either a pure software agent originated by a mobile sink , or a human provided with a special device communicating with the network , or even a mobile robot interacting directly with the network . the case of a software agent is somehow more complicated , because it has to send the results back to the originator of the agent , typically a mobile sink . that is the reason why we focus on the case of a software agent in the present work . each trace can be represented by the following record : .attributes of traces [ cols="^,<",options="header " , ] [ simulation parameters ] the default value value for the density is @xmath1 . in the first experiment set , we test the algorithm with the following densities : @xmath2 , @xmath1 , @xmath3 and @xmath4 , which means that the expected number of neighbours in the communication graph is about @xmath5 , @xmath6 , @xmath7 and @xmath8 respectively , since we use the disc graph model for our simulations ( c.f . section [ sec simuls ] ) , whereas each point of the network can be roughly expected to be sensing covered by @xmath9 to @xmath10 sensor nodes . ( i.e. we choose density parameters which always guarantee that the network is connected with high probability , however the sensing coverage goes from being partial to dense ) . the results are shown on figure [ fig dens ] . in figure [ dist1 ] , we see that the tracker frequently localizes the target ( each time the distance is @xmath11 ) , whatever the density tested except for the lower density . in figure [ messages1 ] , we see that the total number of messages per node is very reasonable for all densities . it is also inter sting to notice that the number of spreading messages seems to be correlated with density . the correlation is not extremely strong ( in reticular above a possible threshold density ) , because other random factors are quite important too ( like the motion pa tern of the target ) . however , we do observe that the number of messages diminishes with the density . this is due to the way the heuristic spanning tree is constructed and is a desirable feature ( specifically the way one of the nodes always tries to go the further possible from the repulsion point , c.f . section [ trace spreading ] ) . [ fig dens ] the default value value for the target speed is @xmath12 km / h . in the second experiment set , we test the algorithm with the following speeds : @xmath13 , @xmath14 , @xmath15 , @xmath16 km / h . the results are shown on figure [ fig speed ] . [ fig speed ] again , we notice that for all the tested target speeds , the target is often localised by the tracker . this is quite nice , since it could have been feared that for the highest target speeds the tracker would never have localized the target . also , the number of messages is kept low . when the target s speed increases , more detection occurs and thus more traces need to be spread , which explains the increased number of messages with increased speed from figure [ messages2 ] . the default value value for the sensing radius is @xmath15 m. in the third experiment set , we test the algorithm with the following sensing radii : @xmath6 , @xmath8 , @xmath17 , @xmath18 . the results are shown on figure [ fig sens ] . [ fig sens ] we observe the same kind of favorable behavior . the target is being exactly localised pretty often , for all sensing radii . for increasing sensing radius there are more detections by the sensors and thus more messages are being generated , but in every case the number of messages is kept very reasonable . the default value value for the message frequency is @xmath19 second , i.e. the tracking agent moves every second and traces are spread every second . in the third experiment set , we test the algorithm with the following message propagation frequencies : @xmath19 , @xmath20 , @xmath13 , @xmath6 , @xmath7 . the results are shown on figure [ fig freq ] . [ fig freq ] the most important observation is that , not surprisingly , is that for long propagation delays ( e.g. when a message is only sent every @xmath7 seconds ) , it is more difficult to localise the target . this is so because with increasing message propagation delay , the tracking agent becomes slower and is thus less efficient at tracking the target . still , except for the longest propagation delays , the target is being localised regularly . a very nice result is that the number of messages sent increases only slowly with reducing propagation delays . that is , if messages are sent twice more often , the total number of messages is far from doubling . this is due to the inhibition mechanism : when messages are sent quickly , the spreading of a trace is more likely to be inhibited by a previous trace which has not yet vanished with time . on the contrary , when propagation delay are long , the spreading of a new trace is likely to span a larger portion of the network , since previous trace intensities will be low due to the time - intensity dissipation implied by our model . therefore , precision can be traded for increased traffic . we used a ` ruby ` implementation to make our simulations . in our simulations , we use the unit disc graph model for building a high level abstratcion of the communication graph associated with the sensor network . the target movement is modelled using the random waypoint model . in this model , the target chooses a random point and moves to it at the chosen speed ( which was a fixed adjustable parameter in our simulations ) . as a consequence , the target moves along line segments . if the target is moving from point @xmath21 to point @xmath22 , starting at time @xmath23 and arriving at time @xmath24 , we used the following model for target detection by the sensor nodes : if a sensor node @xmath25 is at distance @xmath26 of the segment @xmath27 and that @xmath26 is smaller than than the sensing radius @xmath28 ( c.f . table [ simulation parameters ] ) , then the node detects the presence of the target at time @xmath29 , where @xmath30 , where @xmath31 is the only point of the segment @xmath27 such that @xmath32 . this means that an event can occur at times taking value in the real numbers ( or at least their floating point representation in the programming language ) . we feel that this fairly complicated simulation environment , when compared to an approach that would discretize time in rounds , is much more precise . at the cost of complicating the implementation of the simulation platform and putting higher computation loads on its execution , we thus used a fairly detailed simulation environment . although events occur at real number valued times , there is only a finite number of event during each time interval . we use this to store events on an _ scheduled events stack _ ordered by time , and execute the events starting at the top of the stack . for example , when the target starts moving from point @xmath21 to @xmath22 , we first compute all detections by sensor nodes and put them on the scheduled events stack , ordered by time . we then `` execute '' the first detection by picking the scheduled event at the top of the stack . the top event , assuming it is an target detection by a sensor node , implies initializing a trace on the detecting node . because this trace will have to be spread a few seconds later ( depending on the simulation parameters ) , we create a `` spreading '' event and insert it on the scheduled event stack and sort the stack according to the scheduled execution time of the event ( a full sorting is not required , since the stack is always kept sorted , it in fact suffices to introduce the new event at the proper place in the stack ) . summarizing , the executing flow of the simulator is ( 1 ) pick the event at the top of the stack , ( 2 ) execute the event ( e.g initialize a trace ) ( 3 ) determine all future events implied by the execution and insert them on the scheduled events stack . start at point ( 1 ) again . we have proposed a lightweight target tracking protocol for wireless sensor networks . inspired by an analogy with the way a lion tracks down an antelope in the savannah by following it s smell . in our protocol , targets leave traces behind them . the intensity of those traces ( like the odor of the antelope ) decreases with time . the sensor network propagates those traces ( like the wind propagates the smell of the antelope ) using an innovative propagating process aiming at spanning the network with a tree of degree 2 with non overlapping branches . the protocol was evaluate trough extensive simulations under representative network operation regimes . it was shown that the protocol is successful at letting a tracking agent localize the target regularly , and that the message overhead was kept low . more precisely , it was shown that the protocol scales particularly well in terms of network density since the localization process is efficient and the message overhead is low , for all tested densities . we also show that the network is capable of tracking targets moving at different speeds ( from slow to quite fast ) . we have also tested different sensing coverage regimes ( from not well covered to redundantly covered ) , and showed that although traffic increases with increased sensing range , target localisation is achieved for all regimes with reasonable message overhead . finally , an important finding is that diminishing propagation delays ( e.g. by increasing the duty cycle of sensor nodes ) does not augment too much the message overhead because of the trace spreading inhibition mechanism . j. liu , j. liu , j. reich , p. cheung , and f. zhao , distributed group management for track initiation and maintenance in target localization applications , proc . workshop on information processing in sensor networks ( ipsn ) , 2003 .	we study the important problem of tracking moving targets in wireless sensor networks . we try to overcome the limitations of standard state of the art tracking methods based on continuous location tracking , i.e. the high energy dissipation and communication overhead imposed by the active participation of sensors in the tracking process and the low scalability , especially in sparse networks . instead , our approach uses sensors in a passive way : they just record and judiciously spread information about observed target presence in their vicinity ; this information is then used by the ( powerful ) tracking agent to locate the target by just following the traces left at sensors . our protocol is greedy , local , distributed , energy efficient and very successful , in the sense that ( as shown by extensive simulations ) the tracking agent manages to quickly locate and follow the target ; also , we achieve good trade - offs between the energy dissipation and latency . oddfootto evenfootto
the _ fermi _ gamma - ray space telescope hosts two instruments , the large area telescope @xcite and the gamma - ray burst monitor @xcite , which together are capable of measuring the spectral parameters of gamma - ray bursts ( grbs ) across seven decades in energy . since the start of gbm and lat science operations in early august 2008 , emission at energies @xmath4100mev has been detected from ten grbs . these detections were made possible by the lat s greater sensitivity and shorter deadtime ( 26@xmath5s ) compared to previous instruments . prior to _ fermi _ , high - energy gamma - rays from grbs with energies up to 18 gev were observed by the egret instrument on - board the _ compton gamma - ray observatory_. the egret observations suggested three types of high - energy emission : an extrapolation of the low energy spectra to the @xmath4100 mev band ( e.g. , * ? ? ? * ) , an additional spectral component during the prompt emission @xcite and in the case of grb 940217 , a gev afterglow which was detectable for 90 minutes after the trigger @xcite . the redshifts of these events were not determined . recently , reported that grb 080514b which triggered agile at lower energies , was detected by the grid instrument up to 300 mev . a photometric redshift of @xmath6 was reported for this event . in the _ fermi _ era , due to the advanced localization capabilities of the lat and the rapid follow - up by the _ swift _ narrow field instruments @xcite and the ground - based follow - up community , redshifts for five of the ten lat bursts have been measured . these include grb 080916c with @xmath7 , a long burst that has the highest inferred isotropic energy , @xmath8ergs ( 10 kev10 gev ) @xcite , and grb 090510 with @xmath9 @xcite , the second short burst seen by the lat and the first short burst to show definitively an additional hard power - law component in the gev band during the prompt phase @xcite . grb 090902b is a long , fairly intense burst with a redshift of @xmath10 @xcite and fluence of @xmath11erg@xmath12 ( 10 kev10 gev ) over the first 25 seconds of the prompt emission . these data give an isotropic energy @xmath13ergs , comparable to that of grb 080916c . similar to grb 090510 , grb 090902b has a significant additional , hard power - law component that appears during the prompt phase . furthermore , a spectral feature at energies @xmath14kev is evident in the gbm spectrum of grb 090902b that is consistent with an extrapolation of the @xmath4100mev power - law emission down to those energies . in previous analyses , @xcite reported evidence for an additional low - energy spectral component below 20kev for @xmath15% of batse bursts . we report on the observations and analysis of gamma - ray emission from grb 090902b measured by the gbm and lat instruments . in section 2 , we present details of the detections by both instruments and summarize the follow - up observations . in section 3 , we show the light curves of the prompt emission as seen by the various detectors and describe the extended emission found in the lat data out to 1 ks after the trigger . in section 4 , we present the time - resolved spectral analysis of the burst emission during the prompt phase . finally , in section 5 , we discuss the physical interpretation of the gbm and lat data , focusing on the implications of the power - law component for models of grb physics . on 2009 september 2 at 11:05:08.31 ut , the _ fermi _ gamma - ray burst monitor triggered on and localized the bright burst grb 090202b ( trigger 273582310 / 090902462 , * ? ? ? * ) . the burst was within the lat field of view initially at an angle of 51@xmath16 from the boresight . this event was sufficiently bright in the gbm that an autonomous repoint request was made , and the spacecraft began slewing within 10 seconds towards the burst . after @xmath1200 seconds , it had pointed the lat boresight to within a few degrees of the final burst localization . it maintained that pointing until @xmath11 ks post - trigger , when the earth s limb began to enter the lat field - of - view ( fov ) . this burst was detected up to @xmath15 mev by gbm , and emission was significantly detected by the lat , with 39 photons above 1 gev . the highest energy photon had @xmath17 gev and arrived 82 seconds after the gbm trigger ; and the initial analyses detected photons as late as 300 seconds after the trigger @xcite . from the lat data , the burst was localized to r.a.(j2000 ) , dec(j2000 ) = 265.00 , 27.33 with a statistical uncertainty of 0.04@xmath16 ( + @xmath180.1@xmath16 systematic ) , enabling target of opportunity observations to begin @xmath112.5 hours after the trigger with the narrow field instruments on _ swift_. a candidate x - ray afterglow within the lat error circle was detected by the x - ray telescope ( xrt , * ? ? ? this source was confirmed to be fading @xcite , and uvot observations revealed the optical afterglow @xcite . the earliest ground - based optical observations were obtained by rotse - iiia @xmath11.4 hours post trigger @xcite . other detections were reported in the optical @xcite , in the near infrared by grond @xcite and in the radio @xcite . the location of the fading source detected by grond was r.a.(j2000 ) , dec(j2000 ) = @xmath19 , + 271927.1 , 3.3 arcminutes from the lat location @xcite . the afterglow redshift of @xmath10 was measured by @xcite using the gmos spectrograph mounted on the gemini - north telescope . in figure [ fig : prompt light curves ] , we show the gbm and lat light curves in several energy bands . the top three panels show data from the most brightly illuminated nai and bgo detectors of the gbm , and the bottom three panels show the lat data with various event selections . in the bottom panel , the measured photon energies are plotted as a function of time , including the highest energy event ( @xmath20gev ) that arrived 82 seconds after the gbm trigger time , @xmath21 . from the gbm light curves , we see that at energies @xmath221mev the prompt phase ends approximately 25 seconds after @xmath21 . detailed analysis of the gbm data for energies 50300 kev yields a formal t90 duration of 21.9 seconds starting at @xmath23s . by contrast , the lat emission @xmath4100mev clearly continues well after this time range . on time scales longer than the prompt phase , the lat detects emission from grb 090902b as late as 1 ks after the gbm trigger . the spectrum of this emission is consistent with a power - law with photon index @xmath24 , and its flux ( @xmath4100mev ) declines as @xmath25 over the interval ( @xmath26s ) . as we note above , the lat observations are interrupted by entry of the earth s limb into the fov , but analysis of data after @xmath27s , when the source location is again unocculted , shows that any later emission lies below the lat sensitivity ( figure [ fig : gev afterglow ] ) . the upper limit we obtain for data after @xmath27 is consistent with an extrapolation of the @xmath2 decay . similar late time emission for energies @xmath4100mev that extends well beyond the prompt phase has been seen for five earlier bursts by _ fermi _ : grb 080916c @xcite ; grb 090323 @xcite ; grb 090328 @xcite ; grb 090510 , independently seen by agile @xcite and by _ fermi _ @xcite ; and grb 090626 @xcite . spectral analysis was performed using the data from both the gbm and the lat . these analyses include data from the nai detectors 0,1,2,9,10 and bgo detectors , and lat `` transient '' class data , with front- and back - converting events considered separately . the nai data are fit from 8 kev to 1 mev and the bgo from 250 kev to 40 mev using the time tagged event ( tte ) data , which are high time resolution data that allow us to define the time intervals based on the structure of the gbm and lat light curves . the lat data are fit from 100 mev to 200 gev . an effective area correction of 0.9 has been fit to the bgo data to match the model normalizations given by the nai data ; this correction is consistent with the uncertainties in the gbm detector responses . the fits were performed with the spectral analysis software package rmfit ( version 3.1 ) . for further details on the data extraction and spectral analysis procedures see @xcite and @xcite . the time - integrated spectrum of grb 090902b is best modeled by a band function @xcite and a power - law component ( table [ tab : spectral params ] ) . the power - law component significantly improves the fit between 8kev and 200gev both in the time - integrated spectrum and in the individual time intervals where there are sufficient statistics . it is also required when considering only the gbm data ( 8kev40mev ) for the time - integrated spectrum , as its inclusion causes an improvement of @xmath282000 in the cstat statistic over the band function alone . when data below @xmath150 kev are excluded , a power - law component can be neglected in the gbm - only fits . we conclude that this power - law component contributes a significant part of the emission both at low ( @xmath1850kev ) and high ( @xmath4100mev ) energies . figure [ fig : prompt spectrum ] shows the counts and unfolded @xmath29 spectra for a band function with a power - law component fit to the data for interval * b * ( when the low energy excess is most significant ) using the parameters given in table [ tab : spectral params ] . spectral evolution is apparent in the band function component from the changing @xmath30 values throughout the burst , while @xmath31 remains soft until interval * e * when it hardens significantly . @xmath31 is similarly hard in interval * f * , after which the band function component is no longer detected . the hardening of @xmath31 is accompanied by an apparent hardening of the power - law index , @xmath32 , which until interval * e * does not exhibit much variation . however , this is not definitive since the flux is too low to constrain @xmath32 in intervals * e * and * f * separately . a spectral fit of the sum of these two intervals confirms the presence of both a harder @xmath31 and a harder @xmath32 , with a clear statistical preference for the inclusion of the power - law component . an equally good fit is obtained in the combined * e * + * f * interval if this power - law has an exponential cut - off at high energies , with the preferred cut - off energy lying above 2gev . finally , we note that in interval * b * , a marginally better fit is achieved using a model with the additional power - law component having an exponential cut - off at high energies . the improvement is at the @xmath33 level and indicates weak evidence for a cutoff in the second component , placing a lower limit on the cutoff energy in this interval of about 1gev . the _ fermi _ data for grb 090902b show for the first time clear evidence of excess emission both at low energies ( @xmath34kev ) and at high energies ( @xmath4100mev ) , while the band function alone fits data at intermediate energies adequately . these excesses are well - fit by a single power - law component suggesting a common origin . this power - law component accounts for @xmath35 of the total fluence in the 10 kev10 gev range , and its photon index is hard , with a value @xmath36 throughout most of the prompt phase . such a hard component producing the observed excess at low energies is difficult to explain in the context of leptonic models by the usual synchrotron self - compton ( ssc ) mechanisms . in the simplest versions of these models , the peak of the ssc emission is expected to have a much higher energy than the synchrotron peak at mev energies , and the ssc component has a soft tail that is well below the synchrotron flux at lower energies and so would not produce excess emission below @xmath37 kev . hadronic models , either in the form of proton synchrotron radiation @xcite or photohadronic interactions @xcite , can produce a hard component with a similar low energy excess via direct and cascade radiation ( e.g. , synchrotron emission by secondary pairs at low energies ) . however , the total energy release in hadronic models would exceed the observed gamma - ray energy of @xmath38ergs significantly and may pose a challenge for the total energy budget . collimation into a narrow jet may alleviate the energy requirements , since the actual energy release from grb 090902b can be smaller by a jet beaming factor @xmath39 from the apparent isotropic value , where @xmath40 is the bulk lorentz factor of the fireball . from the observation of a @xmath41 gev photon in interval * c * , the highest energy during the prompt phase and thus the most constraining , we derive a minimum value of the bulk lorentz factor @xmath42 using the flux variability time scale of @xmath43ms found in the bgo data . this limit follows from the constraint that the opacity for @xmath44 pair production with target photons fitted by the band+pl model in interval * c * is less than unity for the 11.16gev photon . this high @xmath45 value is of the same order as the values derived for grb 080916c @xcite and grb 090510 @xcite , both of which have been detected at @xmath46gev with the lat . the delayed onset of the @xmath47100 mev emission from the gbm trigger has been modeled for grb 080916c as arising from proton synchrotron radiation in the prompt phase @xcite and for grb 090510 as arising from electron synchrotron radiation in the early afterglow phase @xcite . in order to produce the peak of the lat emission at @xmath48s in the early afterglow scenario for grb 090902b from deceleration of the grb fireball , a value of @xmath49 is required . this is similar to @xmath45 that we calculate , but the observed large amplitude variability on short time scales ( @xmath50 ms ) in the lat data , which is usually attributed to prompt emission , argues against such models . also , the appearance of the power - law component extending down to @xmath51 kev within only a few seconds of the grb trigger disfavors an afterglow interpretation . the proton synchrotron model , on the other hand , requires a rather large total energy budget , as mentioned previously . yet another interpretation of the observed excess in the high and low energies may be provided by two non - thermal power - law components along with a thermal component from the jet photosphere @xcite . the thermal component , broadened by temperature variations , then accounts for the @xmath52kev few mev emission with @xmath53 @xcite , although fits of such a model to our data do not improve over the band+pl model . furthermore , it is difficult for the photospheric model to explain the delayed onset of the @xmath54100mev emission . the detection of the @xmath55gev photon , @xmath56seconds after the grb trigger and well after the soft gamma - ray emission subsided , may help constrain the origin of the late - time decay of the power - law component , which goes as @xmath2 . a synchrotron origin of the 33.4 gev photon would be difficult since it would require significant energy gain by electrons over a gyroradius and a bulk lorentz factor @xmath41500 . in the case of diffusive shock - acceleration , the energy losses in the upstream region of the shock may dominate ( see , e.g. , * ? ? ? * ) and prevent acceleration of electrons to an energy high enough to radiate a 33.4 gev photon . an interpretation by afterglow ssc emission is still possible , however . the constraints on the quantum gravity mass scale from grb 090902b using the time - of - flight test @xcite are much weaker than those from grb 090510 @xcite due to the larger interval , 82 seconds , between @xmath21 and the arrival time of the 33.4gev photon . however , the moderately high redshift ( @xmath0 ) of grb 090902b allows us to use this photon to probe and constrain models of the extragalactic background light . the 33.4 gev photon would not be absorbed by the ebl in any models except for the `` fast evolution '' and the `` baseline '' models by @xcite , which give optical depths of @xmath57 and 5.8 , respectively . we have performed spectral fits of the lat data with and without the predicted ebl absorption from stecker s models assuming a simple power - law as the intrinsic emission model . based on monte - carlo simulations , we found that stecker s fast evolution and baseline models are disfavored at a @xmath58 level . in summary , grb 090902b is one of the brightest bursts detected by the gbm and lat instruments on _ fermi_. it clearly shows excess emission at high and low energies during the prompt phase , requiring a hard power - law component in addition to the usual band function in order to fit the data . the origin of this component is not understood , and its presence in this burst poses genuine challenges for the theoretical models . like the other two bright _ fermi _ bursts detected by the lat , grb 080916c and grb 090510 , grb 090902b appears to possess a very high lorentz factor for the bulk outflow , @xmath59 , and has some suggestion of a delayed onset of the emission above @xmath1100 mev . finally , the 33.4gev photon , the highest energy yet detected from a grb , and the @xmath0 redshift of this burst have allowed us to place significant constraints on some models of the extragalactic background light . the _ fermi _ lat collaboration acknowledges support from a number of agencies and institutes for both development and the operation of the lat as well as scientific data analysis . these include nasa and doe in the united states , cea / irfu and in2p3/cnrs in france , asi and infn in italy , mext , kek , and jaxa in japan , and the k. a. wallenberg foundation , the swedish research council and the national space board in sweden . additional support from inaf in italy and cnes in france for science analysis during the operations phase is also gratefully acknowledged . , b. l. , catelli , j. r. , & schneid , e. j. 1998 , in american institute of physics conference series , vol . 428 , gamma - ray bursts , 4th hunstville symposium , ed . c. a. meegan , r. d. preece , & t. m. koshut , 349353 100mev and @xmath41gev , respectively . the vertical lines indicate the boundaries of the intervals used for the time - resolved spectral analysis . those time boundaries are at @xmath60 seconds . the insets show the counts for the corresponding dataset binned using these intervals in order to illustrate the relative numbers of counts considered in each spectral fit . ] + s ) . top : counts spectrum ; separate model components are plotted , band ( dashed ) , power - law ( solid ) . bottom : unfolded @xmath61 spectrum . the extension of the @xmath62mev power - law component to the lowest energies ( @xmath63kev ) is shown.,title="fig : " ] cccccccccccc * @xmath64 & 0.030.0 & 726 ( @xmath658 ) & -0.61 ( @xmath650.01 ) & -3.8 ( @xmath66 ) & -1.93(@xmath67 ) & 2562/963 & 2005 & ( 4.59 @xmath65 0.05)@xmath6810@xmath69 + * a. & 0.04.6 & 526 ( @xmath6512 ) & -0.09 ( @xmath650.04 ) & -3.7 ( @xmath70 ) & -1.87(@xmath71 ) & 901/963 & 43 & ( 3.72 @xmath65 0.13)@xmath6810@xmath72 + * b. & 4.6 9.6 & 908 ( @xmath73 ) & 0.07 ( @xmath650.03 ) & -3.9 ( @xmath66 ) & -1.94 ( @xmath650.02 ) & 1250/963 & 3165 & ( 1.44 @xmath65 0.03)@xmath74 + * c. & 9.613.0 & 821 ( @xmath6516 ) & -0.26 ( @xmath650.03 ) & -5.0(@xmath75 ) & -1.98 ( @xmath650.02 ) & 1310/963 & 2109 & ( 9.42 @xmath65 0.24)@xmath76 + * d. & 13.019.2 & 529 ( @xmath659 ) & -0.65 ( @xmath65 - 0.02 ) & -3.2 ( @xmath77 ) & -1.86 ( @xmath65 0.02 ) & 1418/963 & 199 & ( 1.29 @xmath650.03)@xmath74 + * e. & 19.222.7 & 317 ( @xmath658 ) & -0.78 ( @xmath65 - 0.02 ) & -2.4 ( @xmath650.1 ) & @xmath64 & 1117/965 & @xmath64 & ( 4.8 @xmath65 0.2 ) @xmath76 + * f. & 22.725.0 & 236 ( @xmath78 ) & -1.30 ( @xmath79 ) & -2.2 ( @xmath650.1 ) & @xmath64 & 1077/965 & @xmath64 & ( 1.0 @xmath65 0.1)@xmath80 + * e.+f . & 19.225.0 & 327 ( @xmath81 ) & -0.91 ( @xmath650.02 ) & -2.6 ( @xmath82 ) & -1.59 ( @xmath83 ) & 1219/963 & 16 & ( 6.1 @xmath650.4)@xmath80 + * g. & 25.030.0 & @xmath64 & @xmath64 & @xmath64 & -1.93 ( @xmath84 ) & 1209/967 & @xmath64 & ( 6.8 @xmath65 0.8)@xmath85 + * * * * * * * * *	we report on the observation of the bright , long gamma - ray burst , grb 090902b , by the gamma - ray burst monitor ( gbm ) and large area telescope ( lat ) instruments on - board the _ fermi _ observatory . this was one of the brightest grbs to have been observed by the lat , which detected several hundred photons during the prompt phase . with a redshift of @xmath0 , this burst is among the most luminous detected by _ fermi_. time - resolved spectral analysis reveals a significant power - law component in the lat data that is distinct from the usual band model emission that is seen in the sub - mev energy range . this power - law component appears to extrapolate from the gev range to the lowest energies and is more intense than the band component both below @xmath150kev and above 100mev . the band component undergoes substantial spectral evolution over the entire course of the burst , while the photon index of the power - law component remains constant for most of the prompt phase , then hardens significantly towards the end . after the prompt phase , power - law emission persists in the lat data as late as 1 ks post - trigger , with its flux declining as @xmath2 . the lat detected a photon with the highest energy so far measured from a grb , @xmath3 gev . this event arrived 82 seconds after the gbm trigger and @xmath150seconds after the prompt phase emission had ended in the gbm band . we discuss the implications of these results for models of grb emission and for constraints on models of the extragalactic background light .
tungsten trioxide , wo@xmath0 , has been extensively studied due to its very attractive and rich properties for technological applications . wo@xmath0 and its derivatives h@xmath5wo@xmath0 and wo@xmath6 are electrochromic,@xcite thermocromic,@xcite and superconducting.@xcite it has been envisaged that wo@xmath0 may become one of the best materials for electrochromic applications such as in energy - efficient windows , smart glasses , antiglare automobile rear - view mirrors , sunroofs , displays , or even tunable photonic crystals@xcite and to reduce photocorrosion.@xcite the wide variety of the underlying electronic instabilities for these properties is mirrored by a multitude of related structural instabilities , which were investigated ever since 1975@xcite and refined later.@xcite all known wo@xmath0 phases are characterized by very large distortions of the archetypal perovskite structure so that even the notion of octahedra tilts is to be taken with some caution . the wo@xmath7 octahedra are so largely distorted that the variance of w o distances in any observed structure is far greater than in most other known perovskite structures.@xcite in this paper we make the fundamental connection between the electronic and structural structure properties of wo@xmath0 and show that most , if not all , structural instabilities can be derived from a careful analysis of its intrinsic cubic phonon instabilities , despite these large deformation amplitudes . the structural properties can be summarized as follows : wo@xmath0 shows no proper melt , surface melting of crystalline material occurs at 1746 k. crystal growth proceeds typically by sublimation and gas transport at temperatures below 1400k . at the highest temperatures the structures is tetragonal ( space group @xmath8 ) with strong antiferrodistortive ( afd ) cation movements so that the wo@xmath7 octahedra are strongly distorted@xcite in an anti - polar pattern . additional rotational octahedral distortions condense in addition to the initial tetragonal displacements when lowering the temperature . they further reduce the symmetry from tetragonal to orthorhombic , monoclinic , triclinic , and finally to a second monoclinic phase . a structural sequence , which contains phases stabilized by temperature , is given in fig.1 . a monoclinic phase ( @xmath9)@xcite and a triclinic phase @xmath10 exist at room temperature.@xcite at higher temperatures , vogt@xcite et al and locherer _ _ et al.__@xcite concluded a transition from @xmath11 to the @xmath12 phase and howard @xcite observed an intermediate @xmath1 phase . locherer and woodward found an additional transition from @xmath12 to @xmath8 near at 980 k to 1200 k. below room temperature , salje et al@xcite reported a transition from the triclinic @xmath10 phase to a polar phase ( @xmath2 ) with no further transitions down to 5k . wo@xmath0 occurs ( almost ) always as oxygen deficient wo@xmath6 with a metal - insulator ( mi ) transition to a metallic phase for high concentrations of oxygen vacancies or doping with alkali metals . superconductivity occurs in the metallic phase@xcite even if the reduced regions are restricted to nano - scale twin boundaries . bulk superconductivity in wo@xmath6 was found in a tetragonal phase with space group @xmath13.@xcite ( bi- ) polaronic electron transport is a characteristic property of wo@xmath6.@xcite as reported by three main experimental sources ( vogt from ref . , locherer from ref . and howard from ref . ) .,width=302 ] numerous first - principles studies were performed on wo@xmath0 in order to characterize its electronic structure ( bulk , thin films and cluster phases@xcite ) the role of oxygen vacancies,@xcite and cation doping.@xcite in this paper we focus on the effect of the electronic structure on the structural stabilities and metastabilities of the various phases . we show that the hybrid functional b1-wc is preferable for the study of the electronic and structural properties of wo@xmath0 over previous approaches . in the first section we check the validity of the b1-wc functional against six known crystallographic phases . the sole disagreement exists for the crystallographic @xmath14 phase ( @xmath2 ) , which yields a higher symmetric @xmath1 structure . in the second section we analyze metastable phases starting from the phonon dispersions of the hypothetic cubic phase , identifying the main phonon instabilities . we then condense various possible combinations of these unstable modes in order to reproduce all experimentally observed structures . this procedure also reveals two new polar phases that are close in energy to the ground - state . as a preliminary test , we performed calculations on wo@xmath0 using several approaches and codes . we tested the norm - conserving pseudopotentials and projected augmented waves with abinit@xcite and vasp@xcite codes working within local density approximation@xcite ( lda ) and different generalized gradient approximations ( gga ) including pbe,@xcite pbesol@xcite and wu - cohen@xcite for the exchange - correlation functional . we also tested the localized orbital approach and the b1-wc@xcite functional as implemented in the crystal code.@xcite as reported by wang _ et al._,@xcite we found that pbesol and wu - cohen gga functionals give good agreement for the structural properties but underestimate the electronic band gap , while the hybrid functionals give good agreement for both the structure and the band gap . wang _ _ et al.__@xcite investigated b3lyp , pbe0 and hse06 hybrid functionals and found that the hse06 gives the best agreement with experiments . in our study , the b1-wc hybrid functional also gives good results , often better compared with the hse06 functional . we note that the hse06 functional was used through the plane wave vasp code , which is highly computer time consuming while the b1-wc with the localized orbital scheme of crystal is more computationally efficient . in what follows , we mainly focus on the results obtained with the b1-wc functional and the crystal code . we have used the all - electron double-@xmath15 basis sets for the oxygen atoms and small core detlev figgen pseudo - potentials,@xcite associated with double-@xmath15 valence basis sets for tungsten . we performed full structural relaxations with a convergence criteria on the root - mean - square of the gradient and displacements smaller than @xmath16 hartree / bohr and @xmath16 bohr respectively . the electronic self - consistent calculations were converged until the difference of the total energy was smaller than 10@xmath17 hartree . the phonon frequencies and born effective charges were computed using frozen phonon numerical differences@xcite and the electric polarization through the berry phase technique.@xcite the integration in the brillouin zone has been performed with a @xmath18 grid of k - points for the cubic unit cell and a @xmath19 grid for cells doubled in the three directions with respect to the cubic one . several dft studies of wo@xmath0 have been performed previously@xcite essentially focusing on the main and most common phases and on the electronic structure analysis with and without oxygen vacancies . a detailed analysis of the complex structural phase diagram of wo@xmath0 is thus missing while a microscopic knowledge of the origin of these different phases would be extremely valuable to understand the unique properties of wo@xmath0 . in this section we start by characterizing the different phases of wo@xmath0 observed experimentally to validate our approach and we will discuss the possible origin of the @xmath2 phase . we will also analyze the electronic structure of these phases and we will discuss how the b1-wc compares with the previous studies . in table [ tab : positions ] we compare our calculated crystallographic data of the @xmath8 , @xmath12 , @xmath11 , @xmath9 , @xmath10 and @xmath1 phases against the experimental measurements . because the @xmath1 phase is not observed experimentally at low temperature , we compare it with the closely related experimental @xmath2 phase . our calculations of the @xmath8 phase are in very good agreement with the observed cell parameters and the atomic positions . the @xmath8 phase consists of highly distorted wo@xmath0 octahedra where the w o bonds dimerize in opposite direction along the [ 110 ] perovskite direction . this w o dimerization forms local dipole - moments that are aligned along the [ 001 ] direction and anti - aligned along the [ 110 ] direction , so that the total dipole - moment cancels . the crystallographic unit cell is elongated along the [ 001 ] direction and compressed along the [ 100 ] and [ 010 ] directions . the @xmath12 phase shows additional octahedra rotations around the @xmath20 axis ( a@xmath21a@xmath21c@xmath22 in the glazer notation@xcite ) , which induces a cell doubling along the [ 001 ] direction . the calculated @xmath23 cell parameter and the @xmath20 component of the atomic positions are in good agreement with experiments while the in - plane displacements are less well reproduced ( tab . [ table1 ] ) . the calculated @xmath24 and @xmath25 cell parameters are smaller than in experiments by 0.1 and the deviation from the tetragonal o@xmath26 position are about two times larger than observed . we clearly overestimate the a@xmath21a@xmath21c@xmath22 distortions . note , however , that our calculations ignore thermal effects while experiments were performed at high temperatures.@xcite a possible comparison would be to extrapolate the experimental data for limited high temperature intervals to 0k but such data are not available for the @xmath12 phase . the orthorhombic @xmath11 phase can be characterized by its additional octahedra rotation about the crystallographic @xmath27 axis ( @xmath28b@xmath29a@xmath21 ) . we find a similar overestimate of the octahedra distortions as for the @xmath12 phase while the calculated cell parameters are underestimated with respect to experiments . we note that the anti - polar distortions along the z axis compares well with experiments for @xmath8 , @xmath12 and @xmath11 . the @xmath9 structure contains an additional octahedra rotation around the crystallographic @xmath30 axis ( a@xmath22b@xmath29c@xmath22 ) . the calculated cell volume is slightly too small ( @xmath310.7% , @xmath320.7% and @xmath321.9% for @xmath24 , @xmath25 and @xmath23 cell parameters respectively ) and the oxygen motions related to the octahedra tilt are overestimated ( table [ table1 ] ) . the @xmath10 phase is similar to the @xmath9 phase if one replaces the in - phase rotation by an out - of - phase rotation with the pattern a@xmath22b@xmath22c@xmath22 . the distortions are anisotropic in all three directions , which causes the cell to be triclinic with the angles @xmath33 , @xmath34 and @xmath35 close to 90@xmath36 . the calculated @xmath24 , @xmath25 and @xmath23 cell parameters deviate from experiments by @xmath310.4% , @xmath321.1% and @xmath320.8% respectively . the largest apparent deviation is in the @xmath14 ( @xmath2 ) phase . the @xmath2 phase is related to the @xmath1 phase through an additional polar distortions along the c axis . while relaxing the low temperature @xmath2 phase we observed that the system always relaxes back in to the higher symmetric @xmath1 structure . et al _ using lda and gga exchange correlation functionals@xcite found a similar effect . in table [ table1 ] we compare our calculated atomic positions and cell parameters of the @xmath1 phase with the experimentally determined @xmath2 phase . the deviations are surprisingly small ( @xmath310.2% , -0.2% and -0.5% for @xmath24 , @xmath25 and @xmath23 lattice parameters ) and even smaller differences for the atomic positions . to further assess the dynamical stability of the @xmath1 phase with respect to a potential @xmath2 ground state , we computed the zone - center phonons and did no observed any unstable mode . the lowest polar mode has a frequency of 158 @xmath37 and is far from being unstable . we also checked whether a soft polar mode can be generated by increasing the cell volume but did not observe any polar instability . we thus follow the argument by wijs _ _ et al.__@xcite that polarity in the experimental study may be stabilized by the presence of oxygen vacancies or by another extrinsic parameter . comparing the structural parameters obtained with other hybrid functionals pbe0 , b3lyp and hse06 reported by wang et al.@xcite , we find close agreement with a smaller error margin for b1-wc . b1-wc gives a much better agreement for the @xmath1 phase with experimental data than using the three hybrid functionals tested by wang et al . : hse06 , b3lyp and pbe0 with errors of @xmath310.6% , @xmath311.3% and @xmath310.2% on @xmath24 , @xmath312.1% , @xmath312.5% and @xmath310.6% on @xmath25 and @xmath310.1% , @xmath313.0% , @xmath311.7% on the @xmath23 parameter . we notice , however , that the b1-wc often underestimates cell parameters while the three other hybrid functionals overestimate the cell parameters of wo@xmath0 . cc [ cols="^,^,^,^,^,^,^",options="header " , ] [ table1 ] in table [ tab : gaps ] we compare the calculated electronic structures for the hypothetical cubic , @xmath8 , @xmath12 , @xmath11 , @xmath9 , @xmath10 and @xmath1 phases and compare them with the experiments and previous dft calculations using pbe0 , hse06 and b3lyp hybrid functionals and gw . for the @xmath9 and @xmath10 phases , experimental data coincide with the b1-wc band gaps . the b1-wc results are similar to those obtained with the hse06 functional while the pbe0 gives a slightly smaller gap energy and b3lyp larger values . the b1-wc band gap is closest to the results of gw calculations , an agreement also observed for the @xmath1 phase . comparing the trend of band gaps between the different phases , we find that the band gap increases from the cubic ( 1.5 ev ) to the lower symmetry phases ( table [ tab : gaps ] ) . this means that both the anti - polar and octahedra tilt distortions increase the band gap of wo@xmath0 . the calculated electronic gap energies is in reasonable agreement with the experimental values for the three low - temperature structures : eg = 2.85 ev for the room temperature monoclinic phase @xmath9 , eg = 2.98 ev for the triclinic phase @xmath10 and eg = 3.28 ev for the monoclinic phase @xmath1 . figure [ fig : dos ] shows the density of states ( dos ) of these three phases to demonstrate their similarity . , @xmath9 and @xmath1 phases with the b1-wc functional.,width=8 ] lccccccccc ' '' '' & ref . & cubic & @xmath8 & @xmath12 & pcnb & p2@xmath38/n & @xmath10 & p2@xmath38/c + b1-wc & & 1.5 & 2.12 & 2.15 & 2.65 & 2.85 & 2.98 & 3.28 + gw & & & & & & 2.90 & 3.00 & 3.30 + pbe0 & & 2.25 & 2.28 & & 3.35 & 3.67 & 3.67 & + hse06 & & 1.67 & 1.71 & & 2.57 & 2.80 & 2.94 & + b3lyb & & 1.89 & 1.85 & & 2.89 & 3.13 & 3.17 & + exp 1 & & & & & & 2.75 & & & + exp 2 & & & & & & 3.21 & 3.25 & + exp 3 & & & 1.75 & & 2.35 & 2.60 & & + the results presented so far give us confidence that the b1-wc functional reproduces well the experimental measurements so that we can now focus on the structural instabilities of the hypothetical @xmath39 cubic parent phase and explain how their condensation give rise to the various known phases of wo@xmath0 . this also allows us to identify novel ferroelectric metastable phases . in each case , we analyze the crystallographic structure through a decomposition of the distortions with respect to the cubic parent phase in terms of symmetry - adapted modes . ( negative frequencies refer to imaginary frequencies , _ i.e. _ to unstable modes ) . the coordinates of the high symmetry points are as follows : @xmath40 ( 0,0,0 ) , x @xmath41 , m @xmath42 and r @xmath43 . thanks to the _ band2eps _ postprocessing script of abinit,@xcite the color of the bands is assigned to each point through the contribution of each atom type to the corresponding eigenvector : red for the tungsten atom and blue for the oxygens.,width=8 ] fig . [ fig : phonon_dispersions ] shows the calculated phonon dispersion curves of hypothetical cubic wo@xmath0 . two branches of instabilities ( imaginary frequencies plotted as negative numbers in fig.[fig : phonon_dispersions ] ) coexist in the brillouin zone . the first unstable branch has its largest imaginary value at @xmath40 . the @xmath40 unstable mode has the irreducible representation ( irrep ) @xmath44 and corresponds to a polar mode . it suggests that the cubic phase of wo@xmath0 is mostly unstable via this polar instability and might be ferroelectric , which we will see later is not exactly the case . the polar instability at @xmath40 propagates toward the x and m points with weak dispersion while it strongly disperses towards the r point . aside from @xmath40 , the modes of this branch are anti - polar . the dispersion of this unstable branch is very similar to the one reported in batio@xmath0 and corresponds to a ferroelectric instability requiring a chain - like correlation of displacements in real space.@xcite the second branch of unstable modes appears between m and r points with smaller amplitudes and a nearly absent dispersion between these two points . the label of the m and r point unstable phonon modes are m@xmath45 and r@xmath46 and they correspond to rotations of the oxygen octahedra . the dispersion of this branch is comparable to what is observed for similar modes in srtio@xmath0 or pbtio@xmath0 and linked to a planar character of the correlations of the atomic displacements in real space . @xcite fig . [ fig : unstable - modes - vesta ] represents a schematic view of the eigenvectors related to the main instabilities of cubic wo@xmath0 . the polar mode at @xmath40 ( @xmath44 , 373i @xmath37 ) shows motion of w against the o atoms , which is the source of a large electrical polarization . the anti - polar modes at the x ( x@xmath47 , 256@xmath48 @xmath37 ) and m points ( m@xmath49 , 147@xmath48 @xmath37 ) are associated to opposite displacements from unit cell to unit cell along the [ 100 ] and [ 110 ] directions respectively . , [ 010 ] and [ 001 ] directions are degenerate . the same applies for the [ 110 ] , [ 101 ] and [ 011 ] directions . ] the m@xmath45 ( 62i @xmath37 ) and r@xmath46 ( 69i @xmath37 ) unstable modes correspond rotations of the oxygen octahedra about the central w atom with consecutive octahedra along the rotation direction moving respectively in the same or opposite directions . using the glazer notation,@xcite the m@xmath45 mode corresponds to @xmath50 and the r@xmath46 mode to @xmath51 . . small red spheres represent the oxygens and large blue spheres represent the tungsten atoms . all the modes are unstable but the x@xmath52 mode , which is discussed in section [ sectionsymmadapt].,width=321 ] starting from the previous unstable modes , we now investigate how their individual and combined condensations in the hypothetical cubic structure give rise to various phases . we then compare their energies and analyze the amplitudes of distortions . we first consider the condensation of unstable @xmath44 , x@xmath47 and m@xmath49 modes . [ fig : energy - gain ] shows the energy gain of the corresponding relaxed phase with respect to the cubic phase . we tested several condensation schemes : ( i ) condensation of the polar @xmath44mode along one ( @xmath53 ) , two ( @xmath54 ) and three ( @xmath3 ) directions ; ( ii ) condensation of the x@xmath47 along one ( @xmath55 ) and two ( @xmath56 ) directions ; ( iii ) condensation of the m@xmath49 mode along one direction ( @xmath8 ) . we observe that the energy gain of the polar instabilities is large and that the @xmath44 polar mode drives a larger gain of energy ( red columns in fig . [ fig : energy - gain ] ) than the anti - polar x@xmath47 and m@xmath49 modes ( green columns in fig . [ fig : energy - gain ] ) . the space group related to the condensation of the m@xmath49 mode corresponds to the high temperature phase observed experimentally ( @xmath8 ) . condensation of the @xmath44 mode along two and three directions produce energy gains larger than its condensation in a single direction so that @xmath44 mode alone will drive the system polar along the [ 111 ] direction with an energy difference between the @xmath54 and @xmath3 phases of 6 mev . we calculated the polarization amplitude in the three @xmath53 , @xmath54 and @xmath3 phases using the berry phase technique and obtain 54 , 69 and 69 @xmath57c.@xmath58 . these polarization values are comparable to those observed in robust ferroelectrics such as pbtio@xmath0 . they can be explained by the opposite motions of w and o atoms , associated with strongly anomalous born effective charges ( 11.73 e for w and -8.78/-1.62 e for o@xmath59/o@xmath60 in good agreement with previous calculations in ref . ) . . red columns are the fe phases arising for the condensation of the polar unstable mode , blue columns are the phases arising from the condensation of the oxygens octahedral rotation unstable modes , green columns are the phases arising from the condensation of anti - polar modes , magenta column represents a phase combining fe and anti - polar modes and cyan columns represent phases combining oxygen octahedral rotations and anti - polar modes . for clarity , the exact value of the energy gain is written in each case.,width=10 ] fig . [ fig : energy - gain ] ( blue columns ) shows the energy gain given by the condensation of the m@xmath45 and r@xmath46 modes along one direction ( @xmath61 and @xmath62 ) , the r@xmath46 mode in three directions ( @xmath63 ) and the orthorhombic @xmath64 phase where the r@xmath46 mode is condensed in two directions and the m@xmath45 mode in one direction ( @xmath65 ) . these distortions lower the energy much less than the polar and antipolar motions . this observation is in line with the modest amplitude of the related phonon instabilities : the energy curvatures at the origin are less negative for the octahedral rotations than for the polar motion . nevertheless the amplitude of octahedral rotations are 10.7 and 11.7 degrees in @xmath61 and @xmath62 , respectively . such large distortions associated to a weak instability highlight relatively small anharmonicities , which might be explained by the absence of @xmath66 cation with respect to regular @xmath66bo@xmath0 perovskites.@xcite beside the @xmath8 phase , none of the previous single irrep mode condensations correspond to observed phases . thus , we now explore the condensation of combined octahedral rotations and polar / anti - polar modes . we depict in fig . [ fig : energy - gain ] the energy gain given by the joint condensation of polar and oxygen rotation modes along three directions ( @xmath4 , in purple color ) and of anti - polar and oxygen rotation modes ( @xmath12 , @xmath11 , @xmath9 , @xmath1 and @xmath10 , in cyan color ) . combining the polar distortion of the low energy @xmath3 phase with additional oxygen rotation modes does not further reduce the energy . instead , it yields a @xmath4 phase slightly higher in energy but with a slightly amplified polarization of 71 @xmath57c.@xmath58 and a slightly reduced oxygen rotation ( with respect to the @xmath63 ) . this emphasizes an unusual competition between these two types of distortions in wo@xmath0 with respect to regular perovskite compounds , where in wo@xmath0 the @xmath4 phase forms a local minimum between the @xmath63 and @xmath3 phases . at the opposite , the mixing of the anti - polar mode m@xmath49 with oxygen rotation modes can drive larger energy gains so that the ferroelectric @xmath3 phase is not the ground state . this is in agreement with experimental observations where the observed phases at all temperatures contain anti - polar motions . amongst investigated phases , the @xmath1 phase appears as the most stable but only marginally , as we observe that the @xmath9 , @xmath1 and @xmath10 phases are all extremely close in energy ( energy gains of 153 , 155 and 155 mev respectively , see fig . [ fig : energy - gain ] ) . consequently , within the precision of our calculations , we can not unambiguously assess which one is the ground state . nevertheless , as discussed in section iii - a and further exemplified in the next section , the @xmath1 phase is in excellent agreement with the experimental @xmath2 ground state , except for a tiny polar distortion . our calculations highlight that , in fact , the @xmath9 and @xmath10 phases observed at higher temperatures are also extremely close in energy . we further notice that the ferroelectric @xmath3 phase , although never observed experimentally , is also relatively close in energy to the ground state ( about 11 mev / f.u . ) . following k. m. rabe , @xcite the non - polar ( or eventually weakly polar in the experimental @xmath2 phase ) ground - state of wo@xmath0 combined with an alternative low - energy ferroelectric phase obtained by polar distortions of the same high - symmetry reference structure makes it a potential antiferroelectric compound . indeed , applying an electric field , it might be possible to open a typical double hysteresis loop from a field - induced first - order transition from the @xmath1 ground state to the @xmath3 polar phase . estimating the critical electric field required to stabilize the @xmath3 phase from @xmath67 , @xcite where @xmath68 is the energy difference between the two phases ( 11.43 mev / f.u . ) , p@xmath69 the spontaneous polarization of the polar phase ( 69@xmath57c.@xmath58 ) and @xmath70 its unit cell volume ( 55 ) , we get the relatively modest value @xmath71 480 kv / cm . for the polar phase @xmath4 we need to apply a greater electric field @xmath71 638 kv / cm to stabilize this phase . this allows us to estimate that the critical field has smillar value with respect to other antiferroelectric material , @xmath71 470kv / cm for zr@xmath72 @xcite and @xmath71 239kv / cm for pbzr@xmath73 . and @xmath74 phase by s. amisi in his http://hdl.handle.net/2268/158512[phd thesis ] ] although this might not be easy to check experimentally on real samples that are typically oxygen deficient and highly conductive , the calculations reveal that stoichiometric wo@xmath0 exhibits all the features of an antiferroelectric compound . to quantify the distortions that appear in the various phases we project the structural distortions onto symmetry adapted modes of the cubic phase using amplimode software.@xcite the results in fig . [ fig : amplimode2 ] show the amplitudes of the modes in the fully relaxed phases from the calculations but non - observed experimentally . in fig . [ fig : amplimode1 ] we show the amplitudes of modes in both the fully relaxed and observed phases , which can be compared . in the following we discuss the competition / cooperation character of the mode distortions . in perovskite oxydes , it is established that the oxygen rotations are in competition with the ferroelectric displacements but less attention has been given to the combinations of other types of mode . often , this cooperation or competition comes from the biquadratic energy term in the free energy expansion with respect to two order parameters . in wo@xmath0 , we observe that the combination between the @xmath44 mode and the r@xmath46 mode along the [ 111 ] direction in the @xmath4 phase has the tendency to reduce the amplitude of the oxygen rotations with respect to the @xmath63 phase ( the r@xmath46 mode is 13% smaller in the @xmath4 phase than in the @xmath63 phase , see fig . [ fig : amplimode2 ] ) while the polar mode is unaffected . as discussed in the previous section , the combination of the @xmath44 and r@xmath46 modes forms a local minimum ( @xmath4 phase ) of higher energy than the @xmath3 phase . this means that the polar distortions are in competition with the oxygen rotations as reported for perovskite oxydes , with the difference that the polar mode amplitude is unaffected and that the @xmath4 phase is locally stable ( the system does not relax into the lowest energy @xmath3 phase ) . the strain can also play an important role , , @xmath3 and @xmath4 phases are 54.01@xmath75 , 50.89@xmath75 , 55.35@xmath75 , 52.71@xmath75 ] but when performing the same calculations at fixed cell parameters ( fixed to the cubic one ) , we find that the @xmath4 phase still forms a local minimum of higher energy than the @xmath3 phase . this unusual energy landscape can come from the marginal gain of energy of the oxygen rotations while large amplitude of rotations are present . on the other hand , the association of the oxgen rotations with the antipolar m@xmath49 mode is cooperative . when we compare the amplitude of the r@xmath46 and m@xmath49 modes of the @xmath8 , @xmath62 and @xmath12 ( fig . [ fig : amplimode2 ] and fig . [ fig : amplimode1 ] ) we find that when both the r@xmath46 and m@xmath49 modes are present together in the @xmath12 phase , their amplitude is slightly higher ( 4% larger ) than when condensed alone ( @xmath8 and @xmath62 phases ) . their combination , however , drives a sizeable gain of energy : the @xmath12 phase is 63 mev and 126 mev lower in energy than the @xmath8 and the @xmath62 phases , respectively . this means that the combination of the oxygen rotations with the antipolar m@xmath49 mode is much more cooperative than the combination with the polar mode @xmath44 . the @xmath11 phase can be understood as a distorted @xmath12 phase with additional m@xmath77 oxygen rotations along [ 010 ] . the resulting tilt pattern is a@xmath21b@xmath29c@xmath22 with a small energy gain of 2 mev with respect to the @xmath12 phase and a reduction of mode amplitudes m@xmath49 , r@xmath46 and m@xmath77 ( 16% , 7% and 22% reduction of the m@xmath49 , r@xmath46 and m@xmath77 modes with respect to the phases where they are condensed alone , _ i.e. _ , @xmath8 , @xmath62 and @xmath78 , respectively ) . the m@xmath77 mode competes with the r@xmath46 and m@xmath49 modes in the sense their combination reduces their amplitude , but they cooperate to lower the energy of the system . in the case of the monoclinc @xmath9 , @xmath1 and triclinic @xmath10 phases , the combination of the m@xmath49 mode with several oxygen rotations ( a@xmath22b@xmath29c@xmath22 for @xmath9 , a@xmath22a@xmath22c@xmath22 for @xmath1 and a@xmath22b@xmath22c@xmath22 for @xmath10 ) lowers the energy of the crystal and with an increase of the mode amplitude with respect to the phases where these modes are condensed independently . for example , the antipolar m@xmath49 mode has his amplitude increased by 11% , 10% and 19% in the @xmath9 , @xmath1 and @xmath10 phases respectively . this means that the dominant r@xmath46 oxygen rotations cooperate with the antipolar m@xmath49 mode to promote the ground state of wo@xmath0 . we note that in the @xmath8 , @xmath53 , @xmath54 , @xmath3 , @xmath62 , @xmath61 , @xmath63 , @xmath12 and @xmath4 phases the mode decomposition shows only the primary modes we have condensed . this is different with the @xmath64 phase in which a secondary mode x@xmath52 appears with a small amplitude in the mode projections while we have condensed only the primary r@xmath46 and m@xmath77 modes ( see fig . [ fig : amplimode2 ] ) . this additional mode appears by improper coupling between the r@xmath46 and m@xmath77 modes such that the symmetry of the @xmath64 structure allows the x@xmath52 mode to develop even though the x@xmath52 mode is not unstable by itself.@xcite similarly , we observe the apparition of several additional secondary modes in the @xmath11 , @xmath9 , @xmath1 and @xmath10 phases , which we discuss in the next section . explored in our study but not observed experimentally.,width=321 ] ; comparison between experiments and our calculations with the b1-wc functional are shown.,width=321 ] the different phases arise from the condensation of one or several unstable modes ( primary modes ) of the cubic parent structure but , in some cases , also include the further condensation of stable modes ( secondary modes ) with significant amplitudes . often , the appearance of such secondary modes of large amplitude can be traced back in their linear coupling with the primary modes.@xcite this can be rationalized from the existence in the energy expansion of the @xmath39 phase of invariant terms of the form @xmath79 where @xmath80 is the amplitude of the secondary mode @xmath81 and @xmath82 the amplitude of the primary mode @xmath48 . depending of the number of primary mode involved , these coupling terms can be bilinear ( pseudo - proper ) , tri - linear ( improper and hybrid - improper ) , quadri - linear , etc . to clarify the origin of secondary modes in several low symmetry phases of wo@xmath0 , we restrict ourself here to the search of such multi - linear invariant terms in the energy expansion around its cubic phase by using the invariants software.@xcite in the last column of table [ tab : mode_decompo_invariants ] we report these linear invariants up to the fourth order obtained for the @xmath64 , @xmath11 , @xmath9 and @xmath1 phases . the letters represent the mode amplitudes ( @xmath83 ) in the directions specified in the third column where the bold letters refer to the primary modes in the structure ( @xmath82 ) and the normal letters to secondary modes ( @xmath84 ) . in the @xmath64 phase , we find that the x@xmath52 mode appears through a trilinear coupling with the oxygen rotations modes r@xmath46 and m@xmath77 ( hybrid - improper coupling , * * ab*c invariant in table [ tab : mode_decompo_invariants ] ) . this additional hybrid - improper x@xmath52 mode is also found in @xmath64 of @xmath66bo@xmath0 perovskites @xcite where the eigenvector corresponds to anti - polar motions of the a cation . in wo@xmath0 the a - cation is absent and x@xmath52 corresponds to similar anti - polar motions but of oxygen instead of the a - site ( see fig . [ fig : unstable - modes - vesta ] ) . in @xmath11 the primary m@xmath49mode condenses along z , the r@xmath46 mode along z and the m@xmath77 mode along y and four additional secondary modes : x@xmath47 and x@xmath52 with a large amplitude and m@xmath52 and m@xmath46 with a small amplitude as well as an additional component of the m@xmath49 mode about the x direction . if we restrict ourself to the strongest x@xmath47 and x@xmath52 modes we find that both are coupled with the r@xmath46 and m@xmath77 modes through a trilinear coupling ( hybrid - improper , * ab*e and * ab*f invariants in table [ tab : mode_decompo_invariants ] ) but also through a quadrilinear coupling with the r@xmath46 mode and the two components of the m@xmath49 mode ( * ad*ce and * ad*cf invariants in table [ tab : mode_decompo_invariants ] ) . we can thus explain the appearance of the x@xmath52 and x@xmath47 modes through a trilinear coupling with the oxygen octahedral rotations and the appearance of the second x component of the anti - polar m@xmath49mode through a coupling with the secondary x@xmath52 and x@xmath47 modes and the primary r@xmath46 mode . the final structure can thus be seen as anti - polar through the m@xmath49 mode along z with a canting of its direction toward the x axis and through the x@xmath47 mode along the y direction , the whole distortions being associated with the @xmath85 pattern of oxygen rotation distortions . the transition from @xmath11 to @xmath9 can be seen as being induced by the condensation of the r@xmath46 mode along the remaining direction for the oxygen rotation octahedral distortions to @xmath85 . this means that we find the same mode coupling as in the @xmath11 phase plus some extra ones due to the additional mode condensation . because we do not induce a new irrep , the couplings are the same ( i.e. trilinear and quadrilinear between the primary r@xmath46 , m@xmath77 and m@xmath49 modes and the secondary x@xmath47 and x@xmath52 modes ) but in different directions from the @xmath11 phase : we observe the x@xmath47 and x@xmath52 modes in two directions instead of one . other modes also appear in the symmetry adapted mode analysis but with smaller amplitudes ( m@xmath46 , m@xmath47 , m@xmath52 and m@xmath86 ) , which we do not include in the invariant analysis . @xmath10 and @xmath1 are very similar in the sense that for both structures we can envisage the condensation of r@xmath46 modes in three directions and the m@xmath49mode in one direction . the difference is that in the @xmath1 phase the r@xmath46 mode is primary with the same amplitude in two directions and a different amplitude in the third direction ( @xmath87where the m@xmath49mode is primary in the z direction ) while in the @xmath10 phase the condensation of the r@xmath46 mode has different amplitudes in three directions ( @xmath88 ) . in @xmath1 the presence of the x@xmath47 and m@xmath47 secondary modes can be explained by trilinear coupling with the r@xmath46 and m@xmath49 primary modes ( * acd+bc*d and * ab*e invariants in table [ tab : mode_decompo_invariants ] ) in a similar way as in the @xmath9 phase . this analysis shows that the low symmetry phases of wo@xmath0 are complex and involve numerous multilinear couplings of modes if one expands the energy about the cubic phase . we note that , amongst possible couplings , the coupling with the secondary x@xmath47 mode is most important in all low symmetry phases . going further , in order to test whether symmetry arguments can lead to the the polar @xmath2 phase using improper - like couplings , we have tested if there exists a coupling with a polar mode at the @xmath40 point . we did not find any couplings with the r@xmath46 , m@xmath77 or m@xmath49 modes . we thus conclude that it is not possible to generate polarity in wo@xmath0 in the limit to these primary modes , which are the ones appearing in other experimental phases . cclccccc & & & & & + & & & & calc . & exp . & + ' '' '' & r@xmath46 * & ( * a,a,0 ) & @xmath89 & 0.61 & & + & m@xmath77 * & ( 0,0,c ) & @xmath78 & 0.38 & & * * ab*c + & x@xmath52 & ( 0,a,0,0,0,0 ) & cmcm & 0.06 & & + ' '' '' & r@xmath46 * & ( 0,0,a ) & @xmath62 & 0.47 & 0.35 & + & * m@xmath77 * & ( 0,b,0 ) & @xmath78 & 0.39 & 0.30 & * * ab*e , * ad*ce + & m@xmath49 * & ( c,0,d ) & @xmath90 & 0.23 & 0.25 & * * ab*f , * ad*cf + & x@xmath47 & ( 0,0,e ,- e,0,0 ) & @xmath55 & 0.22 & 0.25 & + & x@xmath52 & ( 0,0,0,0,f , f ) & @xmath55 & 0.07 & 0.06 & + ' '' '' & r@xmath46 * & ( 0,a,b ) ' '' '' & @xmath91 & 0.60 & 0.48 & * * aeg+bd*f ' '' '' + & m@xmath77 * & ( * c,0,0 ) & @xmath78 & 0.34 & 0.37 & * ach+aci-bch+bc*i + & m@xmath49 * & ( 0,d,e ) & @xmath90 & 0.27 & 0.25 & * * acdg@xmath32bce*f + & x@xmath47 & ( f ,- f,0,0,g , g ) & @xmath92 & 0.09 & 0.26 & * adeh+adei+bdeh@xmath32bde*i + & x@xmath52 & ( 0,0,h , i,0,0 ) & @xmath56 & 0.06 & 0.08 & + ' '' '' & @xmath93 * & ( * a,-a,-b * ) & @xmath94 & 0 & 0.067 & + & * r@xmath46 * & ( -b,a,-a ) & @xmath95 & 0.73 & 0.72 & * * acd+bc*d + & m@xmath49 * & ( * c,0,0 ) & @xmath8 & 0.27 & 0.3 & + & x@xmath47 & ( 0,0,0,-d,0,0 ) & @xmath96 & 0.10 & 0.14 & + in this study , we have performed a first - principles study of wo@xmath0 using the b1-wc hybrid exchange - correlation functional and highlighted that , in comparison to various other functionals , it yields the best overall agreement with experiments regarding electronic and structural properties . starting from the inspection of the phonon dispersion curves of an hypothetical cubic structure taken as reference , we have identified two main branches of instabilities and characterized various phases arising from the condensation of one or more unstable modes . although the dominant phonon instability is associated to a zone - center polar mode , we found a non - polar @xmath1 ground state arising from the combination of cooperative anti - polar distortions and oxygen octahedra rotations . this phase is very similar to the experimentally reported polar @xmath2 ground state , except for the absence of a tiny polar distortion . our calculations does not show however any tendency of the @xmath1 phase to evolve to a @xmath2 phase suggesting that wo@xmath0 is likely not intrinsically ferroelectric . instead the ferroelectric character might arise from extrinsic defects such as oxygen vacancies . the @xmath1 phase is anti - polar and defects could easily produce a slightly unbalanced anti - dipole structure , yielding a weak net polarization . in this sense , off - stoichiometric wo@xmath0 might be better described as a _ ferrielectric _ compound.@xcite the ground state is determined by two antiparallel movements of w off - centerings which exactly compensate each other in the @xmath1 phase . the displacements are almost identical in the @xmath2 phase , but the two displacements do not fully compensate each other . we suspect that such weak ferrielectricity can be induced by defects such as oxygen vacancies . at the level of our calculations , the @xmath1 ground - state is almost degenerated in energy with the @xmath9 and @xmath10 phases observed at higher temperature . also , we discovered the existence of a never observed and low - energy ferroelectric @xmath3 phase with a large polarization . although this might not be of direct interest due to the conductive character of usual off - stoichiometric samples , the proximity with the @xmath1 ground - state of this structurally - unrelated @xmath3 polar phase toward which the system could be switch through a first - order transition under moderate electric fields , makes wo@xmath0 a potential antiferroelectric material . investigations of the influence of polarons on the wo@xmath0 structures are currently underway . we thank julien varignon for his helpful guidance with the crystal code and emilio artacho for interesting discussions . h. h. , e. b. and ph . g. acknowledge the consortium des equipements de calcul intensif ( ceci ) , funded by the frs - fnrs ( grant 2.5020.11 and and no . 1175545 ) , and the prace project thedenomo for computing facilities and the arc project aimed for financial support . e. k. h. s. is grateful for support to epsrc and the leverhulme trust . h. h. thanks the averroes - erasmus mundus project . 90ifxundefined [ 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 \doibase http://dx.doi.org/10.1016/0030-4018(78)90126-8 [ * , ( ) ] link:\doibase 10.1039/b612174h [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.solmat.2007.01.026 [ * * , ( ) ] , \doibase http://dx.doi.org/10.1016/j.solmat.2009.08.021 [ * * , ( ) ] link:\doibase 10.1002/adma.200501953 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.solmat.2006.09.005 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/s0257-8972(97)00154-0 [ * * , ( ) ] , link:\doibase 10.1063/1.120064 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.solmat.2007.01.026 [ * * , ( ) ] , http://stacks.iop.org/0953-8984/10/i=22/a=003 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1080/00150197608241431 [ * * , ( ) ] link:\doibase 10.1007/s100510050969 [ * * , ] link:\doibase 10.1103/physrevlett.89.227002 [ * * , ( ) ] http://scitation.aip.org/content/aip/journal/apl/96/3/10.1063/1.3292587 [ * * , ( ) ] http://www.sciencedirect.com/science/article/pii/s0927024899000884 [ * * , ( ) ] link:\doibase 10.1039/c0jm00604a [ * * , ( ) ] link:\doibase 10.1107/s0567739475000745 [ * * , ( ) ] link:\doibase 10.1107/s0567739475000757 [ * * , ( ) ] http://stacks.iop.org/0953-8984/11/i=21/a=303 [ * * , ( ) ] http://stacks.iop.org/0953-8984/9/i=31/a=010 [ * * , ( ) ] http://stacks.iop.org/0953-8984/11/i=35/a=312 [ * * , ( ) ] link:\doibase 10.1006/jssc.1999.8173 [ * * , ( ) ] http://stacks.iop.org/0953-8984/14/i=3/a=308 [ * * , ( ) ] http://stacks.iop.org/0953-8984/14/i=23/a=314 [ * * , ( ) ] link:\doibase 10.1107/s0567740878005014 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0022-3697(95)00063-1 [ * * , ( ) ] \doibase http://dx.doi.org/10.1006/jssc.1997.7268 [ * * , ( ) ] link:\doibase 10.1143/jpsj.15.573 [ * * , ( ) ] link:\doibase 10.1107/s0365110x66002469 [ * * , ( ) ] \doibase http://dx.doi.org/10.1063/1.1702176 [ * * , ( ) ] \doibase http://dx.doi.org/10.1088/0953-8984/14/3/308 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/10/22/003 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/10/33/002 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0038-1098(80)91164-3 [ * * , ( ) ] http://stacks.iop.org/0022-3719/13/i=36/a=005 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0022-4596(79)90229-9 [ * * , ( ) ] link:\doibase 10.1080/13642818408238882 [ * * , ( ) ] link:\doibase 10.1063/1.3313945 [ * * , ( ) ] http://stacks.iop.org/0953-8984/25/i=20/a=205502 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.165203 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.2684 [ * * , ( ) ] link:\doibase 10.1021/jp992814b [ * * , ( ) ] link:\doibase 10.1103/physrevb.55.7508 [ * * , ( ) ] link:\doibase 10.1021/ja961514u [ * * , ( ) ] link:\doibase 10.1021/jp201057 m [ * * , ( ) ] link:\doibase 10.1063/1.3505688 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.073103 [ * * , ( ) ] link:\doibase 10.1016/j.ssi.2003.08.012 [ * * , ( ) ] , link:\doibase 10.1103/physrevb.70.165110 [ * * , ( ) ] link:\doibase 10.1021/jp4123387 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.2436 [ * * , ( ) ] link:\doibase 10.1103/physrevb.72.075109 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.115118 [ * * , ( ) ] link:\doibase 10.1021/jp963724z [ * * , ( ) ] link:\doibase 10.1103/physrevb.77.195102 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/25/16/165501 [ * * , ( ) ] link:\doibase 10.1016/j.cpc.2009.07.007 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.11169 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.1758 [ * * , ( ) ] link:\doibase 10.1103/physrevb.48.5031 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.3865 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.136406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.235116 [ * * , ( ) ] link:\doibase 10.1103/physrevb.77.165107 [ * * , ( ) ] link:\doibase 10.1103/physrevb.36.891 [ * * , ( ) ] link:\doibase 10.1063/1.3119665 [ * * , ( ) ] link:\doibase 10.1002/jcc.20019 [ * * , ( ) ] link:\doibase 10.1002/jcc.20120 [ * * , ( ) ] link:\doibase 10.1103/physrevb.47.1651 [ * * , ( ) ] link:\doibase 10.1021/jp045655r [ * * , ( ) ] link:\doibase 10.1107/s0567740872007976 [ * * , ( ) ] @noop _ _ , edited by , , and , proceedings of the nato advanced study institute ( , ) p. http://stacks.iop.org/0953-8984/9/i=31/a=010 [ * * , ( ) ] link:\doibase 10.1002/adfm.201002477 [ * * , ( ) ] link:\doibase 10.1103/physrevb.60.836 [ * * , ( ) ] link:\doibase 10.1103/physrevb.56.983 [ * * , ( ) ] link:\doibase 10.1021/jp402046 t [ * * , ( ) ] `` , '' in link:\doibase 10.1002/9783527654864.ch7 [ _ _ ] ( , ) pp . link:\doibase 10.1103/physrevb.90.140103 [ * * , ( ) ] link:\doibase 10.1107/s0021889809028064 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/27/28/283202 [ * * , ( ) ] link:\doibase 10.1038/nature06817 [ * * , ( ) ] link:\doibase 10.1039/c5dt00010f [ * * , ( ) ] link:\doibase 10.1107/s0021889803005946 [ * * , ( ) ] link:\doibase 10.1016/j.jssc.2012.04.012 [ * * , ( ) ] link:\doibase 10.1103/physrev.120.1670 [ * * , ( ) ]	using first - principles calculations , we analyze the structural properties of tungsten trioxide wo@xmath0 . our calculations rely on density functional theory and the use of the b1-wc hybrid functional , which provides very good agreement with experimental data . the hypothetical high - symmetry cubic reference structure combines several ferroelectric and antiferrodistortive ( antipolar cation motions , rotations and tilts of oxygen octahedra ) structural instabilities . the instability related to antipolar w motions combines with those associated to oxygen rotations and tilts to produce the biggest energy reduction , yielding a @xmath1 ground state . this non - polar @xmath1 phase is only different from the experimentally reported @xmath2 ground state by the absence of a very tiny additional ferroelectric distortion . the calculations performed on a stoichiometric compound so suggest that the low temperature phase of wo@xmath0 is not intrinsically ferroelectric and that the reported ferroelectric character might arise from extrinsic defects such as oxygen vacancies . independently , we also identify never observed @xmath3 and @xmath4 ferroelectric phases with large polarizations and low energies close to the @xmath1 ground state , which makes wo@xmath0 a potential antiferroelectric material . the relative stability of various phases is discussed in terms of the couplings between different structural distortions , highlighting a very complex interplay involving improper - like couplings up to fourth order in the energy expansion in the cubic phase .
ultra - luminous infrared galaxies ( ulirgs ) radiate most of their extremely large , quasar - like luminosities ( @xmath2 ) as infrared dust emission , and dominate the bright end of the galaxy luminosity function in the nearby universe @xcite . recent studies have revealed that the bulk of the cosmic sub - mm background emission has been resolved into discrete sources , similar to nearby ulirgs @xcite , and for this reason data on nearby ulirgs have been extensively used to derive information on star - formation rates , dust content , and metallicity in the early universe @xcite . understanding the nature of nearby ulirgs is therefore of particular importance both locally and cosmologically . in contrast to the majority of less infrared luminous ( @xmath3 10@xmath4 ) galaxies , it has been suggested that energetically important , dust - obscured agns are present in ulirgs ( veilleux , kim , & sanders 1999a ; fischer 2000 ) . if the geometry of the dust distribution is toroidal and/or the amount of dust along our line of sight is not too great , signatures of such dust - obscured agns can be found in the optical and/or near - infrared wavelength range at @xmath5 2 @xmath0 m @xcite . however , detecting agns that are deeply embedded in a spherical dust shell ( hereafter buried agns ) and estimating their energetic importance is very difficult at @xmath5 2 @xmath0 m , even though they are energetically important , since spectral tracers are dominated by less strongly obscured star - formation activity . to detect such buried agns and estimate their energetic importance , we must observe at wavelengths where the effects of dust extinction are smaller . at wavelengths 313 @xmath0 m , dust extinction is lower . furthermore , emission is dominated by radiation from dust , the dominant emission mechanism of ulirgs . thus , discussions of the energy sources of ulirgs are more straightforward than those based on observations at x - ray or radio wavelengths , which are also potentially capable of detecting buried agns , but are dominated by emission mechanisms other than dust emission . finally , by using spectral features at 313 @xmath0 m , we can distinguish the energy sources of individual galaxies : while star - formation - dominated galaxies show strong polycyclic aromatic hydrocarbon ( pah ) emission features , buried agns should show dust absorption features below smooth continuum emission @xcite . using spectral features , particularly the 7.7 @xmath0 m pah emission , in the rest - frame 511 @xmath0 m spectra obtained with the _ infrared space observatory _ ( iso ) , it has been argued that the majority of ulirgs are star - formation powered @xcite . however , the determination of the continuum level with respect to which emission and absorption features should be measured in _ iso _ spectra is highly uncertain , due to insufficient wavelength coverage longward of these features . distinguishing between star - formation and buried agn activity therefore remains controversial @xcite . as discussed in @xcite , observations at 34 @xmath0 m have two important advantages : firstly , dust extinction is as small as that at 78 @xmath0 m @xcite ; and secondly , the 3.3 @xmath0 m pah emission and 3.4 @xmath0 m carbonaceous dust absorption features can be used to distinguish between the different energy sources of ulirgs , without serious uncertainty in the continuum determination . ground - based 34 @xmath0 m spectroscopy is thus potentially a powerful tool to investigate the energetic importance of agns buried in the compact nuclei of ulirgs , and to settle the controversy discussed above . our particular interest is the search for buried agns in ulirgs with cool far - infrared colors and/or non - seyfert optical spectra , which are typically taken to be star - formation - dominated , based on _ iso _ studies ( genzel & cesarsky 2000 ; taniguchi et al . 1999 ; lutz , veilleux , & genzel 1999 ) . ugc 5101 is such a ulirg ( table 1 in this paper ; veilleux et al . 1999a ) , and has actually been diagnosed as being star - formation powered @xcite . in this letter , we report on the results of 34 @xmath0 m spectroscopic observations of ugc 5101 , and the discovery of evidence for an energetically important buried agn in this source . throughout this paper , @xmath6 @xmath7 75 km s@xmath8 mpc@xmath8 , @xmath9 = 0.3 , and @xmath10 = 0.7 are adopted . we used the nsfcam grism mode @xcite to obtain a 34 @xmath0 m spectrum of ugc 5101 at irtf on mauna kea , hawaii on the night of 2001 april 9 ( ut ) . sky conditions were photometric throughout the observations , and the seeing was measured to be 0@xmath1170@xmath118 in full - width at half maximum . the detector was a 256 @xmath12 256 insb array . the hkl grism and l blocker were used with the 4-pixel slit (= 1@xmath13 ) . the resulting spectral resolution was @xmath14150 at 3.5 @xmath0 m . the spectrum of ugc 5101 was obtained toward the flux peak at 34 @xmath0 m . the position angle of the slit was 0 degree east of north . a standard telescope nodding technique with a throw of 12@xmath15 was employed along the slit to subtract background emission . hr 4112 ( f8v , v=4.8 ) was observed with an airmass difference of @xmath3 0.1 to correct for the transmission of the earth s atmosphere , and provide flux calibration . standard data analysis procedures were employed , using iraf . initially , bad pixels were replaced with the interpolated signals of the surrounding pixels . bias was subsequently subtracted from the obtained frames and the frames were divided by a spectroscopic flat frame . finally the spectra of ugc 5101 and hr 4112 were extracted . wavelength calibration was performed taking into account the wavelength - dependent transmission of the earth s atmosphere . since we set each exposure time to 1.2 sec for the ugc 5101 observation , to reduce observing overheads , data at @xmath16 4 @xmath0 m were affected by the non - linear response of the detector . we excluded these data from our analysis . the ugc 5101 spectrum was then divided by the observed spectrum of hr 4112 and multiplied by the spectrum of a blackbody with a temperature appropriate to f8v stars ( 6200k ) . by adopting @xmath17 = 3.4 for hr 4112 based on @xmath18 = 1.4 @xcite , a flux - calibrated spectrum was produced . a flux - calibrated 34 @xmath0 m spectrum of ugc 5101 is shown in figure [ fig1 ] . at 34 @xmath0 m , ugc 5101 was spatially very compact , and its spatial extent along the slit direction was indistinguishable from a point source . the spatially - unresolved red near - infrared nucleus reported by @xcite and interpreted as a possible agn by these authors is the probable origin of the bulk of the detected 34 @xmath0 m continuum in our spectrum . our spectrum gives a value of @xmath17 ( 3.55 @xmath0 m ) of 10.1 mag , which is similar to the value of @xmath19 ( 3.7 @xmath0 m ) of 9.8 mag measured with a 5@xmath15 aperture @xcite . given that ugc 5101 has a red near - infrared color ( @xmath20 = 1.3 ; sanders et al . 1988a ) , so that @xmath21 should be @xmath16 0 mag , our slit loss for the continuum emission is less than 0.3 mag . the dust continuum emission at 1020 @xmath0 m @xcite , pa@xmath22 emission ( genzel et al . 1998 , the top - middle of fig 8) , and 618 cm radio continuum emission @xcite all show a centrally peaked morphology , with spatially - extended emission presumably originating in weakly obscured star - formation activity . our 1@xmath112 slit covers the bulk of this emission , so that most of the pah emission should also be covered with our slit . any missing pah flux is very likely to be much smaller than the detected pah flux , and will not seriously affect our quantitative discussion of the energy source of ugc 5101 ( @xmath23 4.1 ) . the spectrum in fig . [ fig1 ] shows a remarkable deviation at 3.33.7 @xmath0 m from the smooth continuum emission . ( a linear continuum level is shown as the solid line in fig . [ fig1 ] . ) we attribute the emission feature to 3.3 @xmath0 m pah emission , since the flux peak at 3.42 @xmath0 m is consistent with the expected peak wavelength of the redshifted pah emission ( 3.29 @xmath0 m @xmath12 1.040 ) . based on our adopted linear continuum level , the observed flux and rest - frame equivalent width of the 3.3 @xmath0 m pah emission are estimated to be f@xmath24 = 1.3 @xmath12 10@xmath25 ergs s@xmath8 @xmath26 and ew@xmath24 = 0.025 @xmath0 m , respectively . at wavelengths longward of the pah emission , signals at 3.53.7 @xmath0 m are suppressed compared to the continuum level . the flux level suddenly and steeply decreases at @xmath143.7 @xmath0 m , compared to the extrapolation from longer wavelengths , so that a strong absorption feature is undoubtedly present . we attribute this flux suppression to the 3.4 @xmath0 m carbonaceous dust absorption @xcite , because the observed maximum of absorption at @xmath143.53 @xmath0 m is consistent with the redshifted absorption peak wavelength of the carbonaceous dust absorption feature ( 3.4 @xmath0 m @xmath12 1.040 ) . its observed optical depth is @xmath27(observed ) @xmath14 0.7 . the 3.3 @xmath0 m pah emission profile almost fades out at 3.333.34 @xmath0 m @xcite , while the 3.4 @xmath0 m carbonaceous dust absorption profile is just beginning at this wavelength range @xcite . at _ z _ = 0.040 , this wavelength range , where neither emission nor absorption is particularly important , is redshifted to 3.463.47 @xmath0 m . it will be seen that in fig . [ fig1 ] the adopted linear continuum intersects the observed data points at this wavelength , providing further evidence that the continuum determination is reasonable . an alternative hypothesis is represented by the curved continuum shown as the dashed line in fig . if this model were adopted , the observed flux and rest frame equivalent width of the 3.3 @xmath0 m pah emission would increase to f@xmath24 = 2.0 @xmath12 10@xmath25 ergs s@xmath8 @xmath26 and ew@xmath24 = 0.045 @xmath0 m , respectively , and the observed optical depth of the 3.4 @xmath0 m dust absorption would decrease to @xmath27(observed ) @xmath14 0.6 . however , the level of this curved continuum seems to be too low ; it requires that the absorption optical depth be close to zero at 3.48 @xmath0 m or 3.35 @xmath0 m in the rest frame , where the optical depth should in fact be significant @xcite . we thus believe that the actual pah flux and equivalent width are lower than the values inferred using the curved continuum level ; we can use these latter values as stringent upper limits . considering the uncertainty of the continuum determination , we combine the values measured based on the two continuum levels and adopt f@xmath24 = [email protected] @xmath12 10@xmath25 ergs s@xmath8 @xmath26 , ew@xmath24 = [email protected] @xmath0 m , and @xmath27(observed ) = [email protected] . the detection of 3.3 @xmath0 m pah emission indicates the presence of star - formation activity . however , its observed luminosity is [email protected] @xmath12 10@xmath29 ergs s@xmath8 , which yields an observed 3.3 @xmath0 m pah to far - infrared luminosity ratio ( table 1 ) of @xmath141 @xmath12 10@xmath30 , an order of magnitude smaller than is found in star - formation - dominated galaxies ( @xmath141 @xmath12 10@xmath31 ; mouri et al . 1990 ) . the detected , weakly - obscured star - formation activity thus contributes little to the huge far - infrared luminosity of ugc 5101 and a dominant energy source must be located behind the dust . the presence of such a dust - obscured energy source is supported by the detection of strong 3.4 @xmath0 m carbonaceous dust absorption . if the 34 @xmath0 m continuum emission source behind the dust originated in star - formation activity , the rest - frame equivalent width of the 3.3 @xmath0 m pah emission should be similar to those of less obscured star - formation - dominated galaxies ( @xmath140.12 @xmath0 m ; imanishi & dudley 2000 ) , because both 34 @xmath0 m continuum and 3.3 @xmath0 m pah emission fluxes would be attenuated similarly by dust extinction . the observed rest - frame equivalent width is , however , smaller by more than a factor of three . consequently , the emission from behind the dust shows virtually no pah emission , suggesting that the source is an agn . this agn activity dominates the observed 34 @xmath0 m continuum flux , contributing @xmath1470 % of it , and is plausibly the energy source of ugc 5101 s far - infrared luminosity . if the dust obscuring such an energetically important agn had a torus - like geometry , we would expect ugc 5101 to be optically classified as a seyfert 2 , but in fact the non - seyfert optical classification of ugc 5101 @xcite implies that the agn is obscured by dust along all lines of sight ( that is , a buried agn ) . after subtracting the contribution from star - formation to the observed 34 @xmath0 m fluxes by using the spectral shape of the starburst galaxy ngc 253 in @xcite , we estimate the intrinsic optical depth of the 3.4 @xmath0 m carbonaceous dust absorption toward the buried agn to be @xmath27(intrinsic ) @xmath14 0.8 , which yields dust extinction of @xmath32 @xmath16 100 mag if a galactic extinction curve is assumed @xcite . assuming a@xmath33 @xmath14 0.05 @xmath12 a@xmath34 @xcite , we estimate the dereddened 34 @xmath0 m dust emission luminosity powered by the buried agn to be @xmath35l@xmath36 @xmath3710@xmath38 ergs s@xmath8 , which is comparable to the _ observed _ peak dust emission luminosity at 60 @xmath0 m and 100 @xmath0 m ( @xmath35l@xmath36 @xmath14 2 @xmath12 10@xmath38 ergs s@xmath8 , as calculated based on the _ iras _ fluxes shown in table 1 ) . in the case of a buried agn , dust radiative transfer controls the temperature of the dust shell ; the temperature of the dust decreases with increasing distance from the central agn . the entire luminosity is transferred at each temperature . the present data thus provide evidence for an inner @xmath14900 k , 34 @xmath0 m continuum emitting dust shell of a luminosity similar to that of the observed outer @xmath1440 k , 60100 @xmath0 m continuum emitting dust shell , as is expected from the buried agn scenario . for obscured agns , @xcite found that the column density of x - ray absorbing gas , parameterized by @xmath39 , relative to dust extinction towards the 34 @xmath0 m continuum emitting region ( @xmath32 ) is often higher by a large factor than the galactic @xmath39/@xmath32 ratio ( 1.8 @xmath12 10@xmath40 cm@xmath8 mag@xmath8 ; predehl & schmitt 1995 ) . dust obscuration toward the 34 @xmath0 m continuum emitting region around the buried agn in ugc 5101 is so high ( @xmath42 100 mag : this work ) that @xmath39 could easily exceed 10@xmath43 @xmath26 , in which case direct 210 kev x - ray emission from the buried agns would be very strongly attenuated . the non - detection of 210 kev x - rays from the putative buried agn with asca @xcite can thus be explained without difficulty . the radio to far - infrared flux ratio of ugc 5101 is a factor of two larger than the typical values for star - formation - dominated galaxies at 1.5 ghz @xcite , but inside their scatter at 151 mhz and 5 ghz @xcite . though the bulk of the radio emission is extended @xcite , the spatially - unresolved vlbi radio core , interpreted as an agn by @xcite , is luminous enough to account for the bolomotric luminosity of ugc 5101 with agn activity @xcite . besides ugc 5101 , studies in the thermal infrared have provided evidence for the presence of energetically dominant buried agns in two other optically non - seyfert ulirgs , iras 08572 + 3915 @xcite and iras f00183@xmath17111 @xcite . the far - infrared emission properties of these three ulirgs are summarized in table 1 . in the optical , all are classified as liners @xcite . for ulirgs classified optically as liners , which constitute @xmath1440% of ulirgs @xcite , two arguments based on _ iso _ studies have been made : ( 1 ) star - formation activity is energetically dominant , and ( 2 ) the liner - type optical line emission is due to superwind - driven shocks caused by star - formation activity @xcite . the first argument , however , is not applicable to the three ulirgs in table 1 . then what is the origin of the liner - type optical emission in these objects ? in buried agns , uv emission is blocked at the inner surface of the surrounding dust shell , but x - rays can penetrate deeply into the dust to produce x - ray dissociation regions ( xdrs ; maloney , holllenbach , & tielens 1996 ) . in xdrs , locally - generated uv photons , such as lyman - werner band h@xmath44 emission and ly@xmath22 emission , can ionize some species with ionization potential less than 1011 ev ( notably carbon and iron ) , but the emitting volume is dominated by largely neutral gas , which causes xdrs essentially always to have liner - type spectra ( maloney , in preparation ) . a buried agn can thus explain the observed liner - type optical emission . however , in practical cases , some star - formation activity may very well be present at the surface of galaxies even when buried agns are energetically dominant , as we infer for ugc 5101 . in such a case , due to the susceptibility of optical emission to dust extinction , the optical spectral diagnostics are likely to be dominated by this surface star - formation emission . the star - formation activity could be responsible for liner - type optical emission via shocks , or might even alter the optical emission to a hii region type . it is important to note that narrow - line emission from high - excitation clouds characteristic of seyfert galaxy spectra which are observed to be produced along the axes of nuclear dust tori would not contribute substantially at any wavelength in the case of buried agn . we have found evidence for the presence of an energetically dominant , buried agn in the ulirg ugc 5101 , which is characterized by a cool far - infrared color and a liner optical spectrum . this finding seems contrary to the currently established view , based on _ iso _ data , that such objects are powered by star formation . two other liner ulirgs also show evidence for energetically dominant buried agns , and we therefore conclude that at least some fraction of liner ulirgs are powered by buried agns . we are grateful to j. rayner and b. golisch for their support before and during the observing run . research in infrared astronomy at nrl is supported by the office of naval research . prm is supported by the nsf under grant ast-9900871 . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . alonso - herrero , a. , ward , m. j. , & kotilainen , j. k. 1997 , mnras , 288 , 977 barger , a. j. , cowie , l. l. , & richards , e. a. 2000 , aj , 119 , 2092 blain , a. w. , kneib , j. -p . , ivison , r. j. , & smail , i. 1999a , apj , 512 , l87 blain , a. w. , smail , i. , ivison , r. j. , & kneib , j. -p . 1999b , mnras , 302 , 632 cox , m. j. , eales , s. a. e. , alexander , p. , & fitt , a. j. 1988 , mnras , 235 , 1227 crawford , t. , marr , j. , partridge , b. , & strauss , m. a. 1996 , apj , 460 , 225 dudley , c. c. 1999 , mnras , 307 , 553 dudley , c. c. , & wynn - williams , c. g. 1997 , apj , 488 , 720 fischer , j. 2000 , astro - ph/0009395 genzel , r. , & cesarsky , c. j. 2000 , ara&a , 38 , 761 genzel , r. et al . 1998 , apj , 498 , 579 imanishi , m. , & dudley , c. c. 2000 , apj , 545 , 701 lonsdale , c. j. , smith , h. e. , & lonsdale , c. j. 1995 , apj , 438 , 632 lutz , d. et al . 1996 , a&a , 315 , l269 lutz , d. , veilleux , s. , & genzel , r. 1999 , apj , 517 , l13 maloney , p. , hollenbach , d. , & tielens , a. g. g. m. 1996 , apj , 466 , 561 mouri , h. , kawara , k. , taniguchi , y. , & nishida , m. 1990 , apj , 356 , l39 nakagawa , t. , kii , t. , fujimoto , r. , miyazaki , t. , inoue , h. , ogasaka , y. , arnaud , k. , & kawabe , r. 1999 , galaxy interactions at low and high redshift , ( eds . ) barnes , j. e. , and sanders , d. b. , iau , 186 , 341 pendleton , y. j. , sandford , s. a. , allamandola , l. j. , tielens , a. g. g. m. , & sellgren , k. 1994 , apj , 437 , 683 predehl , p. , & schmitt , j. h. m. m. 1995 , a&a , 293 , 889 rigopoulou , d. , spoon , h. w. w. , genzel , r. , lutz , d. , moorwood , a. f. m. , & tran , q. d. 1999 , aj , 118 , 2625 roche , p. f. , aitken , d. k. , smith , c. h. , & ward , m. j. 1991 , mnras , 248 , 606 sanders , d. b. , & mirabel , i. f. 1996 , ara&a , 34 , 749 sanders , d. b. , soifer , b. t. , elias , j. h. , madore , b. f. , matthews , k. , neugebauer , g. , & scoville , n. z. 1988a , apj , 325 , 74 sanders , d. b. , soifer , b. t. , elias , j. h. , neugebauer , g. , & matthews , k. 1988b , apj , 328 , l35 scoville , n. z. et al . 2000 , aj , 119 , 991 shure , m. a. , toomey , d. w. , rayner , j. t. , onaka , p. , & denault , a. j. 1994 , proc . spie , 2198 , 614 smith , h. e. , lonsdale , c. j. , & lonsdale , c. j. 1998 , apj , 492 , 137 soifer , b. t. , sanders , d. b. , madore , b. f. , neugebauer , g. , danielson , g. e. , elias , j. h. , lonsdale , c. j. , & rice , w. l. 1987 , apj , 320 , 238 soifer et al . 2000 , apj , 119 , 509 sopp , h. m. , & alexander , p. 1991 , mnras , 251 , 112 spoon , h. w. w. , keane , j. v. , tielens , a. g. g. m. , lutz , d. , & moorwood , a. f. m. 2001 , a&a , 365 , l353 taniguchi , y. , yoshino , a. , ohyama , y. , & nishiura , s. 1999 , apj , 514 , 660 tokunaga , a. t. 2000 , in allen s astrophysical quantities , ed . a. n. cox ( 4th ed : aip press : springer ) , chapter 7 , p.143 tokunaga a. t. , sellgren k. , smith r. g. , nagata t. , sakata a. , nakada y. , 1991 , apj , 380 , 452 tran , q. d. et al . 2001 , apj , 552 , 527 veilleux , s. , kim , d. -c . , & sanders , d. b. 1999a , apj , 522 , 113 veilleux , s. , sanders , d. b. , & kim , d. -c . 1999b , apj , 522 , 139 lcrrrcl ugc 5101 & 0.040 & 1.03 & 11.54 & 20.23 & 45.61 & 0.09 ( cool ) + iras 08572 + 3915 & 0.058 & 1.70 & 7.43 & 4.59 & 45.62 & 0.23 ( warm ) + iras f00183@xmath17111 & 0.327 & 0.13 & 1.20 & 1.19 & 46.51 & 0.11 ( cool ) +	we report on the results of 34 @xmath0 m spectroscopy of the ultra - luminous infrared galaxy ( ulirg ) ugc 5101 . it has a cool far - infrared color and a liner - type optical spectrum , and so , based on a view gaining some currency , would be regarded as dominated by star formation . however , we find that it has strong 3.4 @xmath0 m carbonaceous dust absorption , low - equivalent - width 3.3 @xmath0 m polycyclic aromatic hydrocarbon ( pah ) emission , and a small 3.3 @xmath0 m pah to far - infrared luminosity ratio . this favors an alternative scenario , in which an energetically dominant agn is present behind obscuring dust . the agn is plausibly obscured along all lines of sight ( a ` buried agn ' ) , rather than merely obscured along our particular line of sight . such buried agns have previously been found in thermal infrared studies of the ulirgs iras 08572 + 3915 and iras f00183@xmath17111 , both classified optically as liners . we argue that buried agns can produce liner - type optical spectra , and that at least some fraction of liner - type ulirgs are predominantly powered by buried agns .
the origin of the black holes ( bhs ) inhabiting the centers of massive galaxies ( e.g. gebhardt et al . 2000 ; merritt & ferrarese 2001 ) and powering luminous quasars at high redshift ( e.g. willott et al . 2003 ; fan et al . 2006 ; mortlock et al . 2011 ) has long been an open question at the forefront of cosmology and galaxy formation . there is a strong possibility that many of these grew from seed bhs which were born with masses of 10@xmath2 - 10@xmath3 m@xmath4 in the centers of atomic cooling dark matter ( dm ) halos in the early universe ( e.g. volonteri 2012 ; natarajan & volonteri 2012 ) . in this scenario , the primordial gas within the halo is unable to cool below @xmath0 10@xmath2 k ( e.g. spaans & silk 2006 ) because of a low abundance of h@xmath1 molecules , which implies that runaway gravitational collapse only occurs once up to @xmath0 10@xmath3 m@xmath4 of gas has accumulated in the center of the halo . the central objects that form from this collapse , likely short - lived supermassive stars ( e.g. fuller et al . 1986 ; hosokawa et al . 2013 ) or quasi - stars ( e.g. begelman et al . 2008 ) , grow at rates of up to 1 m@xmath4 yr@xmath5 ( wise et al . 2008 ; regan & haehnelt 2009 ; shang et al . 2010 ; johnson et al . 2011 ; latif et al . 2013a ; prieto et al . 2013 ) and are believed to typically leave behind bhs with masses of up to 10@xmath3 m@xmath4 ( e.g. choi et al . bhs formed via this process are referred to as direct collapse black holes ( dcbhs ) . in dcbh formation , the h@xmath1 fraction is typically considered to be suppressed by a strong molecule - dissociating , lyman - werner ( lw ) radiation field ( e.g. haiman 2000 ; glover & brand 2001 ; machacek et al . 2001 ; ahn et al . 2012 ) . while an elevated lw radiation field is likely required to sufficiently suppress h@xmath1 cooling during the collapse of the gas ( e.g. bromm & loeb 2003 ; shang et al . 2010 ; van borm & spaans 2013 ; visbal et al . 2014a ) , the results of both semi - analytic models ( e.g. dijkstra et al . 2008 ; petri et al . 2012 ; agarwal et al . 2012 ; fernandez et al . 2014 ; ferrara et al . 2014 ; visbal et al . 2014b ) and cosmological simulations ( agarwal et al . 2014 ) suggest that such fields may have been produced regularly in the early universe and therefore that dcbh formation may have been relatively common . this conclusion has been strengthened by complementary modeling of the growth ( e.g. johnson et al . 2012 ; latif et al . 2013b ; regan et al . 2013 ) and evolution ( e.g. begelman 2010 ; hosokawa et al . 2012 , 2013 ; inayoshi et al . 2013 ; schleicher et al . 2013 ) of the supermassive stellar progenitors of dcbhs , which suggests that their growth to mass scales of up to @xmath0 10@xmath3 m@xmath4 is not likely impeded by radiative feedback or pulsational instability . indeed , there is a strong possibility that dcbh remnants still reside in present - day galaxies ( koushiappas et al . 2004 ; bellovary et al . 2011 ; devecchi et al . 2012 ; greene 2012 ; reines et al . 2014 ) , including our milky way ( rashkov & madau 2014 ) and its satellites ( van wassenhove et al . 2010 ) while the elevated lw radiation fields required for widespread dcbh formation may have been produced in the early universe , it is likely that in many cases other forms of radiation accompanied them . in particular , given that the sources of this radiation were likely early star forming galaxies , ionizing radiation from stars , x - rays from young accreting bhs and cosmic rays from early supernovae may have commonly been present , in addition to lw radiation . using one zone models of primordial gas collapse in atomically cooled haloes , inayoshi & omukai ( 2011 ) addressed the role that x - rays and cosmic ray ionization may have played in the process of dcbh formation , finding that the free electron population generated deep in the cores of the haloes could have catalyzed the rapid formation of h@xmath1 molecules , leading to cooling of the primordial gas to temperatures well below 10@xmath6 k and preventing dcbh formation . in addition , yue et al . ( 2014 ) have recently developed a semi - analytic model in which dcbh formation is halted once the universe becomes reionized , due to the photoevaporation of the gas from atomic cooling haloes . here , we address the effects of ionizing radiation on the formation of dcbhs using cosmological simulations that track both the elevated lw radiation field required for dcbh formation as well as an accompanying ionizing radiation field . in section 2 we describe the methodology used to model the effects of background ionizing and lw radiation fields . in section 3 we present the results of our simulations . finally , in section 4 , we give a brief discussion of our results . here we describe the approach that we have taken to model the impact of a photoionizing background on the process of dcbh formation . we have carried out two cosmological simulations employing the same version of the smoothed particle hydrodynamics ( sph ) code gadget ( springel et al . 2001 ; springel & herquist 2002 ) that we have employed in previous work ( see e.g. johnson et al . 2011 , 2013 ) . we make use of the same initial conditions as in those previous works , namely a 1 mpc@xmath6 ( comoving ) cosmological volume which is initialized at @xmath7 = 100 and within which an atomic cooling halo is identified at @xmath7 @xmath8 15 , at which time its dm mass is @xmath8 4 @xmath9 10@xmath10 m@xmath4 ( corresponding roughly to a 3-@xmath11 fluctuation ) . it is the evolution of the primordial gas during its collapse into this halo that is the focus of our study . in both of these simulations we use the same prescription for the lw background radiation field that we have employed in the previous works cited above . this consists of a constant uniform background lw field with an intensity characterized by @xmath12 = 10@xmath6 , with a spectrum charaterized by a temperature of @xmath0 10@xmath2 k ( shang et al . units of 10@xmath13 erg s@xmath5 @xmath14 hz@xmath5 sr@xmath5 . ] we account for the self - shielding of h@xmath1 molecules to this radiation by calculating the h@xmath1 self - shielding factor , which expresses the fraction of the unattenuated background lw flux to which a gas parcel is exposed , using an estimate based on the local column density of h@xmath1 molecules ( bromm & loeb 2003 ; see also shang et al . 2010 ; wolcott - green et al . 2011 ) . along with this constant ( in time ) and uniform lw radiation field , we model the effect of an accompanying constant uniform ionizing radiation field . = 100 in our simulations , the ionizing background is turned on at @xmath7 = 30 in our fiducial case . in appendix b , we discuss the impact of turning the ionizing background on later , at @xmath7 = 20 . ] we adopt the following value for the local photoionization rate , assuming a ratio of ionizing photons to lw photons appropriate for a low - metallicity stellar population with an age of @xmath0 10@xmath10 yr ( leitherer et al . 1999 ) : @xmath15 where @xmath16 is the local optical depth to ionizing photons , estimated as described below . the value this obtains for the unattenuated photoionization rate ( with @xmath16 = 0 ) is chosen to be consistent with our choice of @xmath12 = 10@xmath6 for the unattenuated lw flux , assuming a value for the ratio of the escape fraction of ionizing photons to the escape fraction of lw photons for the galaxies producing the radiation of @xmath17/@xmath18 @xmath0 0.3 , which is in broad agreement with estimates gleaned from simulations of early dwarf galaxies ( e.g. ricotti et al . 2001 ; kitayama et al . 2004 ; wise & cen 2009 ; razoumov & sommer - larsen 2010 ; paardekooper et al . 10@xmath10 yr , then the photoionization rate we have adopted will be an overestimate . in particular , the rate , normalized to the lw flux , may be roughly two orders of magnitude lower for a population age of @xmath0 10@xmath19 yr ( e.g. leitherer et al . 1999 ) , although we emphasize that in the early universe ( e.g. at @xmath7 @xmath20 6 ) the stars producing the bulk of the lw flux are likely much younger than this . ] for the corresponding photoheating rate , we conservatively assume that @xmath8 2 ev is deposited in the gas as heat for each photoionization . we adopt a local approximation for the flux of ionizing photons to which a given parcel of gas is exposed , by estimating the optical depth to ionizing photons as @xmath21 where @xmath22 @xmath8 6 @xmath9 10@xmath23 @xmath14 ( @xmath24/13.6 ev)@xmath25 is the cross section for photoionization of neutral hydrogen , = 15.6 ev , consistent with the 2 ev that we assume is deposited in the gas for each photoionization . ] @xmath26 is the local density of neutral hydrogen atoms and @xmath27 is the physical length scale appropriate for the parcel of gas , which we take to be defined in terms of the mass @xmath28=120 m@xmath4 of an sph particle in our simulation : @xmath29 where @xmath30 is the gas density at the location of the sph particle . note that we have normalized equations ( 2 ) and ( 3 ) for a neutral primordial gas ; in particular , the optical depth can be @xmath16 @xmath31 1 where the gas is highly ionized . while this method provides a simple approximation for the local ionization and heating rates , as discussed further in appendix a , we expect that our local estimate for the optical depth to ionizing photons ( equation 2 ) is valid . this is supported by the fact that the results of our simulations for the gas density and collapse redshift required for self - shielding from ionizing radiation are in very good agreement with the detailed estimates recently presented by noh & mcquinn ( 2014 ; see their equations 6 and 7 ) . our results are also consistent with those found by dijkstra et al . ( 2004 ) for the conditions required for the gas in high redshift haloes to self - shield against an intergalactic photoionizing background . these authors found that the gas self - shields more readily at high redshift due to the higher densities in virialized haloes . the atomic cooling halo we focus on in our simulations , and into which gas does collapse in the presence of an ionizing backgroud , has a circular velocity of @xmath8 20 km s@xmath5 at @xmath7 @xmath8 15 , which is higher than the minimum circular velocity that dijkstra et al . ( 2004 ) find is required to retain gas in the presence of a photoionizing background at this redshift . furthermore , this is also consistent with the results of okamoto et al . ( 2008 ) , who found from cosmological simulations that haloes with circular velocities @xmath20 10 km s@xmath5 are able to retain a large fraction of their baryonic mass at @xmath7 @xmath20 10 . this agreement with previous work gives us confidence that our approach produces reliable results . @xmath8 10@xmath2 cm@xmath25 , at the same time as shown in the top panel of figure 1 ( @xmath7 = 14.2 ) . clockwise from top - left : the number density of hydrogen atoms , the h@xmath1 fraction , the free electron fraction and the temperature . contours denote the distribution of the gas , with the mass fraction varying by an order of magnitude across contour lines . the high free electron fraction in the photoionized gas leads to efficient h@xmath1 formation within the central @xmath0 100 pc where the gas is self - shielded to the ionizing radiation . in turn , this leads to efficient cooling of the gas by h@xmath1 , despite the elevated molecule - dissociating lw radiation field , and the temperature falls to 200 k. it is unlikely that a dcbh can form from gas at such low temperatures.,width=316 ] = 15.2 ) . in this case , the free electron fraction in the gas passsing through the virial shock never becomes larger than @xmath0 10@xmath33 and , consequently , the h@xmath1 fraction remains below @xmath0 10@xmath34 . at such a low abundance h@xmath1 cooling does not operate efficiently and the gas stays at temperatures high enough for a dcbh to form.,width=316 ] here we compare the results of our two simulations , highlighting the effects of the ionizing background on the evolution of the primordial gas . as we shall see , this radiation can profoundly alter the final outcome of the collapse of the gas . figure 1 shows the properties of the gas within a 10 pc ( physical ) slice of the simulation volume , centered on the atomic cooling halo , when the gas at the center of the halo has collapsed to a maximum density of @xmath26 @xmath8 10@xmath2 cm@xmath25 . this corresponds to redshifts @xmath7 = 14.2 and 15.2 , for the simulations with and without ionizing radiation , respectively . the effect of the photoheating in the low density intergalactic medium ( igm ) is evident in the top panels , which show that the gas is evaporated out of the cosmological filaments that feed the central halo ; as shown in the bottom panels , these structures are intact in the simulation with only lw radiation . this photoheating and evaporation of the filamentary gas results in a higher gas pressure and a slower accretion rate of gas into the halo , as is shown by the lower ( negative ) infall velocities of the gas in the panels on the right . this results in the gas collapsing to @xmath26 @xmath8 10@xmath2 cm@xmath25 roughly 25 myr later in the simulation including ionizing radiation . figures 2 and 3 show the properties of the gas in the central halo , as functions of the distance from the densest gas particle . in contrast to the case with just lw radiation , in the case with ionizing radiation , at large distances the gas is highly ionized and is almost entirely heated to temperatures up to @xmath8 10@xmath2 k. at distances @xmath35 100 pc , the gas is self - shielded to the ionizing radiation and can collapse to high densities . due to the higher free electron fraction of the collapsing gas in the case with ionizing radiation , at high densities the formation of h@xmath1 is catalyzed ( via the same reactions through which h@xmath1 is generally formed in the primordial gas ; see e.g. bromm & larson 2004 ) and a much larger fraction of h@xmath1 is generated than in the case with just lw radiation . due to the higher h@xmath1 fraction in the case with ionizing radiation , the gas is able to cool via molecular transitions down to @xmath0 200 k , as is typical for population ( pop ) iii star formation in minihaloes ( see e.g. greif et al . 2011 ; clark et al . these temperatures are much lower than the @xmath0 10@xmath6 10@xmath2 k to which the gas can cool in the case with lw radiation only . the corresponding accretion rates onto the objects that form via the runaway gravitational collapse of the gas at the center of the halo are expected to be very different , because of the large difference in the gas temperature ( e.g. omukai & palla 2003 ) : @xmath36 using the temperatures at the center of the halo shown in figs . 2 and 3 , for the cases with and without ionizing radiation , this corresponds roughly to @xmath8 10@xmath25 m@xmath4 yr@xmath5 and @xmath8 10@xmath5 m@xmath4 yr@xmath5 , respectively . again , the former is consistent with pop iii star formation in which h@xmath1 cooling is effective , while the latter is consistent instead with the formation of a supermassive star ( or binary supermassive stars ; see e.g. regan & haehnelt 2009 ; whalen et al . 2013 ) with a mass of @xmath0 10@xmath37 m@xmath4 ( e.g. wise et al . 2008 ; shang et al . 2010 ; johnson et al . 2012 ; latif et al . 2013b ) . from this , we can conclude that when there is only lw radiation it is likely that a supermassive star ( and subsequently a dcbh ) will form , while in the case with both lw and ionizing radiation a small galaxy composed of pop iii stars with masses @xmath35 10@xmath6 m@xmath4 will likely form instead ( e.g. hirano et al . 2014 ) . figure 4 shows the enclosed mass of h@xmath1 , as a function of distance from the center of the halo , in both simulations . the much higher h@xmath1 mass in the case with ionizing radiation , leads to strong self - shielding of the gas to lw radiation , as shown in figure 5 . this effective self - shielding in turn implies that the h@xmath1 photodissociation rate in the center of the halo is much lower than in the case with just lw radiation . this leads to a higher h@xmath1 fraction , which in turn leads to stronger self - shielding of the gas . thus , there is a runaway process of h@xmath1 formation and self - shielding , set up by the fact that the gas collapses into the halo with an elevated free electron fraction because it was photoionized in the igm . the slower formation of h@xmath1 in the case with just lw radiation leads to weaker self - shielding and , ultimately , to the weak molecular cooling and higher temperatures that set the stage for the formation of a dcbh . figure 6 shows both the mass enclosed and the radially - averaged jeans mass of the gas , as a function of the distance from the center of the halo , for both simulations . runaway gravitational collapse is possible when the enclosed mass exceeds the jeans mass . as shown in figures 6 and 7 , there is a factor of a few less gas in the center of the halo in the case with ionizing radiation . however , due to the lower temperatures of the gas in this case , we expect that the central @xmath35 10@xmath2 m@xmath4 of gas becomes jeans unstable and undergoes runaway gravitational collapse , while in the case with just lw radiation the gas is jeans unstable at a much larger mass scale of @xmath0 10@xmath37 m@xmath4 . this is again consistent with our expectation that a supermassive star , and subsequently a dcbh , will form in the case with just lw radiation , while a small pop iii galaxy will instead form in the case including ionizing radiation . molecules enclosed , as a function of distance from the center of the atomic cooling halo , in our two simulations at the same redshifts as shown in the previous figures . due to the elevated free electron fraction in the case with photoionization , a larger mass of h@xmath1 builds up than in the case with just lw radiation . this leads to the gas becoming self - shielding to lw radiation , as shown in figure 5.,width=326 ] -dissociating lw flux is decreased due to local self - shielding of the h@xmath1 molecules , as a function of distance from the center of the host halo . shown are the radially - averaged values of this factor , for lw and photoionizing background radiation fields ( _ blue _ ) and for just a lw background radiation field ( _ orange _ ) , at the same times as shown in the previous figures . the elevated h@xmath1 fraction in the simulation including ionizing radiation leads to much stronger self - shielding of the molecules , which allows for efficient formation of h@xmath1 . , width=326 ] , which roughly corresponds to the minimum mass that can be resolved . the gas is only able to collapse under its own gravity when the enclosed mass is larger than the jeans mass . in the case with lw and ionization this occurs at a mass scale of a few @xmath9 10@xmath6 m@xmath4 , whereas in the case with only lw radiation this occurs at a mass scale of @xmath0 10@xmath37 m@xmath4 . in the former case , the mass scale is too small for the formation of a dcbh , whereas it is sufficiently large for a dcbh to form in the latter.,width=326 ] @xmath8 500 pc is lower by a factor of @xmath8 3 . note that , as shown in figure 6 , the mass within @xmath38 @xmath8 2 pc is not well - resolved in our simulation.,width=326 ] we have presented a pair of cosmological simulations which demonstrate that it is possible for ionizing radiation to prevent dcbh formation in atomic cooling haloes . we find that this is due to the large free electron fraction in the photoionized gas , which leads to the rapid formation of h@xmath39 and ultimately to effective molecular cooling that sterilizes the halo for dcbh formation . we expect that a small pop iii galaxy with up to @xmath0 10@xmath2 m@xmath4 in primordial stars may form , instead , similar to previously studied cases of pop iii star formation in atomic cooling haloes including just a background lw radiation field ( e.g. oh & haiman 2002 ; trenti & stiavelli 2009 ; safranek - shrader et al . 2012 ) . we emphasize that this effect relies on the gas being photoionized before its collapse into the center of the halo . in cases where the ionizing radiation turns on at relatively late times , when the gas in the halo has already collapsed to densities at which it is self - shielded from the ionizing radiation , we do not expect that a large h@xmath1 fraction develops , and indeed it is likely that dcbh formation is still able to take place in this case . in appendix b , we present one such case , in which the ionizing background turns on at @xmath7 = 20 , by which time the dense gas in the halo is already self - shielded from the ionizing radiation . indeed , we find that in this case the collapse of the self - shielded gas proceeds almost identically to the case with just lw radiation ( compare figures 3 and b1 ) . because many atomic cooling haloes may be self - shielding to ionizing radiation , especially in the early stages of reionization ( see e.g. noh & mcquinn 2014 ) , dcbh formation may still take place readily at @xmath7 @xmath20 15 even in photoionized regions . we note that this picture is in basic agreement with that presented by yue et al . ( 2014 ) , who argue that in photoionized regions the gas can not be retained by atomic cooling haloes at @xmath7 @xmath35 14 , and so dcbhs will not likely form in photoionized regions below this redshift . our result is related , but distinct we find that if the gas in atomically cooling haloes is subject to an elevated ionizing radiation background and can not self - shield against it , then dcbh formation can be prevented _ even if the gas is retained in the halo_. we also emphasize that , because reionization is an inhomogeneous process , it should in principle be possible for dcbhs to form at redshifts @xmath7 @xmath35 14 in regions which are not yet reionized ; consistent with this , agarwal et al . ( 2014 ) find that dcbhs may form down to at least @xmath7 @xmath0 9 . our results suggest that , when ionizing radiation accompanies a background lw radiation field the critical lw flux required for dcbh formation is likely to be significantly higher than in the absence of ionizing radiation . as we adopted @xmath12 = 10@xmath6 for our simulations , it appears that the critical flux for photoionized gas may be well above this value , although this is likely to vary with redshift and may be a function of the growth history of the halo ( see e.g. latif et al . this correction could lead to reduced estimates of the prevalence of dcbhs in the early universe , as compared to previous results ( see e.g. dijkstra et al . 2008 ; shang et al . 2010 ; agarwal et al . 2012 , 2014 ; latif et al . 2014 ) , at least in reionized regions . related to the question of the critical lw flux , we note that here we have followed the approach of similar works ( e.g. bromm & loeb 2003 ; shang et al . 2010 ) and adopted the approximation given by draine & bertoldi ( 1996 ) for the h@xmath1 self - shielding factor . as shown by shang et al . ( 2010 ) and wolcott - green et al . ( 2011 ) , this likely overestimates the shielding factor , and so may lead to overestimates of the critical lw flux required to suppress molecular cooling ( but see also richings et al . therefore , the critical flux for the halo in our simulations may in fact be significantly lower than the @xmath12 = 10@xmath6 that we adopted . as we have used the same self - shielding prescription in our simulations both with and without ionizing radiation , we do not expect that using an improved prescription would qualitatively change our central results pertaining to the effects of ionizing radiation . that said , it may be crucial to model in great detail both the photodissociation of h@xmath1 molecules and the transfer of the h@xmath1 line emission that can cool the gas ( e.g. greif 2014 ) , in order to determine the final fates of atomic cooling haloes exposed to elevated levels of lw radiation . we expect that the pop iii galaxies that form in cases in which ionizing radiation prevents dcbh formation may contain significantly more mass in primordial stars than the stellar clusters formed in smaller minihaloes . this follows from the fact that the relatively deep dm gravitational potential wells of atomic cooling haloes allow the gas to be more readily retained in the face of stellar feedback than in the case of pop iii star formation in minihaloes ( see e.g. bromm & yoshida 2011 ) . due to their higher masses , we expect that pop iii galaxies formed in atomic cooling haloes may be among the most luminous pop iii star - forming objects , with distinct observational signatures that could be detected by future missions such as the _ james webb space telescope _ ( see e.g. schaerer 2003 ; johnson 2010 ; inoue 2011 ; zackrisson et al . 2011 ; pawlik et al . work at lanl was done under the auspices of the national nuclear security administration of the us department of energy at los alamos national laboratory under contract no . de - ac52 - 06na25396 . all simulations were performed on the institutional computing ( ic ) network platform mustang at lanl . d.j.w . was supported by the european research council under the european community s seventh framework programme ( fp7/2007 - 2013 ) via the erc advanced grant `` starlight : formation of the first stars '' ( project number 339177 ) . we would like to thank joseph smidt , john wise and hui li for valuable discussions which helped to shape this work . 99 agarwal , b. , khochfar , s. , johnson , j. l. , neistein , e. , dalla vecchia , c. , livio , m. 2012 , mnras , 425 , 2854 agarwal , b. , dalla vecchia , c. , johnson , j. l. , khochfar , s. , paardekooper , j .- 2014 , mnras , submitted ( arxiv:1403.5267 ) ahn , k. , iliev , i. , shapiro , p. r. , mellema , g. , koda , j. , mao , y 2012 , apj , 756 , 16 begelman m. c. 2010 , mnras , 402 , 673 begelman m. c. , rossi , e. m. , armitage , p. j. 2008 , mnras , 370 , 289 bellovary , j. , volonteri , m. , governato , f. , shen , s. , quinn , t. , wadsley , j. 2011 , apj , 742 , 13 bromm , v. , loeb , a. 2003 , apj , 596 , 34 bromm , v. , yoshida , n. 2011 , ara&a , 49 , 373 bromm , v. , larson , r. b. 2004 , ara&a , 42 , 79 chen , k .- j . , heger , a. , woosley , s. , almgren , a. , whalen , d. j. , johnson , j. l. 2014 , apj , submitted ( arxiv:1402.4777 ) choi , j .- h . , shlosman , i. , begelman , m. c. 2013 , apj , submitted ( arxiv:1304.1369 ) clark , p. c. , glover , s. c. o. , smith , r. j. , greift , t. h. , klessen , r. s. , bromm , v. 2011 , sci , 331 , 1040 devecchi , b. , volonteri , m. , rossi , e. m. , colpi , m. , zwart , s. p. 2012 , mnras , 421 , 1465 dijkstra , m. , haiman , z. , rees , m. j. , weinberg , d. h. 2004 , apj , 601 , 666 dijkstra , m. , haiman , z. , mesinger , a. , wyithe , j. s. b. 2008 , mnras , 391 , 1961 dijkstra , m. , ferrara , a. , mesinger , a. 2014 , mnras , 442 , 2036 draine , b. t. , bertoldi , f. 1996 , apj , 468 , 269 fan , x. , et al . 2006 , aj , 131 , 1203 fernandez , r. , bryan , g. l. , haiman , z. , li , m. 2014 , mnras , submitted ( arxiv:1401.5803 ) ferrara , a. , salvadori , s. , yue , b. , schleicher , d. r. g. 2014 , mnras , submitted ( arxiv:1406.6685 ) fryer , c. l. , heger , a. 2011 , an , 332 , 408 fuller , g. m. , woosley , s. e. , weaver , t. a. 1986 , 307 , 675 gebhardt , k. , et al . 2000 , apj , 539 , l13 glover , s. c. o. , brand , p. w. j. l. 2001 , mnras , 321 , 385 greene , j. e. 2012 , nat , in press ( arxiv:1211.7082 ) greif t. h. 2014 , mnras , submitted ( arxiv:1404.5311 ) greif t. h. , springel , v. , white , s. d. m. , glover , s. c. o. , clark , p. c. , smith , r. j. , klessen , r. s. , bromm , v. 2011 , apj , 737 , 75 haiman , z. , abel , t. , rees , m. j. 2000 , apj , 534 , 11 hirano , s. , hosokawa , t. , yoshida , n. , umeda , h. , omukai , k. , chiaki , g. , yorke , h. w. 2014 , apj , 781 , 60 hosokawa , t. , omukai , k. , yorke , h. w. 2012 , apj , submitted ( arxiv:1203.2613 ) hosokawa , t. , yorke , h. w. , inayoshi , k. , omukai , k. , yoshida , n. 2013 , apj , submitted ( arxiv:1308.4457 ) inayoshi , k. , hosokawa , t. , omukai , k. 2013 , mnras , accepted ( arxiv:1302.6065 ) inayoshi , k. , omukai , k. 2011 , mnras , 416 , 2748 inoue , a. k. 2011 , mnras , 415 , 2920 johnson , j. l. , khochfar , s. , greif , t. h. , durier , f. 2011 , mnras , 410 , 919 johnson , j. l. , whalen , d. j. , even , w. , fryer , c. l. , heger , a. , smidt , j. , chen , k .- j . 2013 , apj , 775 , 107 johnson , j. l. , whalen , d. j. , fryer , c. l. , li , h. 2012 , apj , 750 , 66 johnson , j. l. 2010 , mnras , 404 , 1425 kitayama , t. , yoshida , n. , susa , h. , umemura , m. 2004 , apj , 613 , 631 koushiappas , s. m. , bullock , j. s. , dekel , a. 2004 , mnras , 354 , 292 latif , m. a. , schleicher , d. r. g. , schmidt , w. , niemeyer , j. 2013a , mnras , submitted ( arxiv:1304.0962 ) latif , m. a. , schleicher , d. r. g. , schmidt , w. , niemeyer , j. 2013b , mnras , 436 , 2989 latif , m. a. , bovino , s. , van borm , c. , grassi , t. , schleicher , d. r. g. , spaans , m. 2014 , mnras , submitted ( arxiv:1404.5773 ) leitherer , c. , et al . 1999 , apjs , 123 , 3 lodato , g. , natarajan , p. 2006 , mnras , 371 , 1813 astropart . phys . , 31 , 376 machacek , m. e. , bryan , g. l. , abel , t. 2001 , apj , 548 , 509 merritt , d. , ferrarese , l. 2001 , apj , 547 , 140 mesinger , a. , bryan , g. l. , haiman , z. 2006 , apj , 648 , 835 montero , p. j. , janka , h .- t . , m " uller , e. 2012 , apj , 749 , 37 mortlock , d. j. , et al . 2011 , nat , 474 , 616 natarajan , p. , volonteri , m. 2012 , mnras , 422 , 2051 noh , y. , mcquinn , m. 2014 , apj , submitted ( arxiv:1401.0737 ) oh , s. p. , haiman , z. 2002 , apj , 569 , 558 okamoto , t. , gao , l. , theuns , t. 2008 , mnras , 390 , 920 omukai , k. , palla , f. 2003 , apj , 589 , 677 paardekooper , j. p. , khochfar , s. , dalla vecchia , c. 2013 , mnras , 429 , l94 pawlik , a. h. , milosavljevi ' c , m. , bromm , v. 2013 , apj , 767 , 59 petri , a. , ferrara , a. , salvaterra , r. 2012 , mnras , 422 , 1690 prieto , j. , jimenez , r. , haiman , z. 2013 , mnras , 436 , 2301 rashkov , v. , madau , p. 2014 , apj , 780 , 187 razoumov , a. o. , sommer - larsen , j. 2010 , apj , 710 , 1239 regan j. a. , haehnelt m. g. 2009 , mnras , 396 , 343 regan j. a. , johansson , p. h , haehnelt m. g. 2013 , mnras , submitted ( arxiv:1312.4962 ) reines , a. e. , plotkin , r. m. , russell , t. d. , mezcua , m. , condon , j. j. , sivakoff , g. r. , johnson , k. e. 2014 , apj , accepted ( arxiv:1405.0278 ) richings , a. j. , schaye , j. , oppenheimer , b. d. 2014 , mnras , 442 , 2780 ricotti , m. , gnedin , n. y. , shull , m. j. 2001 , apj , 560 , 580 safranek - shrader , c. , agarwal , m. , federrath , c. , dubey , a. , milosavljevi ' c , m. , bromm , v. 2012 , mnras , 426 , 1159 schaerer , d. 2003 , a&a , 397 , 527 schleicher , d. r. g. , palla , f. , ferrara , a. , galli , d. , latif , m. 2013 , a&a , submitted ( arxiv:1305.5923 ) sethi , s. , haiman , z. , pandey , k. 2010 , apj , 721 , 615 shang , c. , bryan , g. l. , haiman , z. 2010 , mnras , 402 , 1249 springel , v. , yoshida , n. , white , s. d. m. 2001 , newa , 6 , 79 springel , v. , hernquist , l. 2002 , mnras , 333 , 649 spaans , m. , silk , j. 2006 , apj , 652 , 902 van wassenhove , s. , volonteri , m. , walker , m. g. , gair , j. r. 2010 , mnras , 408 , 1139 van borm , c. , spaans , m. 2013 , a&a , submitted ( arxiv:1304.4057 ) visbal , e. , haiman , z. , bryan , g. l. 2014a , mnras , submitted ( arxiv:1403.1293 ) visbal , e. , haiman , z. , bryan , g. l. 2014b , mnras , submitted ( arxiv:1406.7020 ) volonteri , m. 2012 , sci , 337 , 544 volonteri , m. , begelman , m. c. 2010 , mnras , 409 , 1022 whalen , d. j. , et al . 2013 , apj , 778 , 17 willott , c. j. , mclure , r. j. , jarvis , m. j. 2003 , apj , 587 , l15 wise , j. h. , turk , m. j. , abel , t. 2008 , apj , 682 , 745 wise , j. h. , cen , r. 2009 , apj , 693 , 984 wolcott - green , j. , haiman , z. , bryan , g. l. 2011 , mnras , 418 , 838 yue , b. , ferrara , a. , salvaterra , r. , xu , y. , chen , x. 2013 , mnras , accepted ( arxiv:1305.5177 ) zackrisson , e. , inoue , a. k. , rydberg , c .- e . , duval , f. 2011 , mnras , 418 , l104 our approach to modelling the impact of an ionizing background radiation field , as described in section 2 , relies on a simple , local estimate of its attenuation . in particular , it is assumed that the attenuation of the field largely takes place over a length - scale comparable to , or smaller than , that of an individual sph particle . here we show that this assumption is valid . the optical depth to ionizing radiation over the length - scale of a parcel of gas represented by a single sph particle , under the assumption of a neutral medium , is given by equation ( 2 ) . evaluated at a photon energy of 15.6 ev ( as we assume in our calculations ; see section 2 ) and expressing this length - scale following equation ( 3 ) , the optical depth becomes @xmath40 as shown in figures 2 and 3 , the density profile of the atomic cooling halo in our simulations can be approximated as isothermal , with the number density of hydrogen nuclei approximated as @xmath26 @xmath8 10@xmath37 ( @xmath38/pc)@xmath33 . using this expression in the above formula for optical depth , we obtain @xmath41 where here @xmath38 is the distance from the center of the halo . this implies that in neutral regions the optical depth is large , i.e. that @xmath16 @xmath20 1 , over the length - scale of an sph particle at @xmath38 @xmath35 10@xmath2 pc . therefore , we expect our approximation to be valid in the dense central regions of the halo , at @xmath38 @xmath0 100 pc , within which we find the gas to be optically thick to the external ionizing radiation field . here we briefly highlight the results of a simulation in which the ionizing background turns on at @xmath7 = 20 , in order to highlight the dependence of our results on the state of the halo when first subjected to the ionizing background . by this redshift , the dense gas in the halo is already self - shielding to the ionizing radiation and it is never photoionized . we note that the gas becoming self - shielded already by @xmath7 = 20 in this halo is consistent with the recent analytical estimates presented by noh & mcquinn ( 2014 ) , as well as with the results of previous numerical simulations cited in section 2 ( dijkstra et al . 2004 ; okamoto et al . 2008 ) . figure b1 shows the properties of the gas , as functions of the distance from the densest sph particle in the halo , just before it undergoes runaway gravitational collapse . comparing this to figure 3 , we see that the gas evolves in much the same way as in the case with just lw radiation . in particular , the free electron fraction and the h@xmath1 fraction in the inner regions of the halo are comparably low in both cases , and the gas remains sufficiently hot for dcbh formation to proceed . this supports our conclusion that dcbh formation is prevented by ionizing radiation only in cases in which the dense gas in the core of the halo is not self - shielding to the radiation and undergoes runaway gravitational collapse only after being photoionized . = 20 . in this case , the gas in the halo has already collapsed to relatively high density by this redshift and the dense core of the halo is self - shielding to the ionizing radiation . as a result , the free electron fraction in the dense gas is not elevated relative to the case with just lw radiation and the h@xmath1 fraction also remains low ( compare to figure 3 ) . therefore , in this case and in others where the dense gas in the core of the halo is self - shielded to the ionizing radiation and never becomes photoionized , we expect that dcbh formation is not prevented and likely proceeds just as in the case with only lw radiation.,width=316 ]	direct collapse black holes ( dcbhs ) formed from the collapse of atomically - cooled primordial gas in the early universe are strong candidates for the seeds of supermassive bhs . dcbhs are thought to form in atomic cooling haloes in the presence of a strong molecule - dissociating , lyman - werner ( lw ) radiation field . given that star forming galaxies are likely to be the source of the lw radiation in this scenario , ionizing radiation from these galaxies may accompany the lw radiation . we present cosmological simulations resolving the collapse of primordial gas into an atomic cooling halo , including the effects of both lw and ionizing radiation . we find that in cases where the gas is not self - shielded from the ionizing radiation , the collapse can be delayed by @xmath0 25 myr . when the ionized gas does collapse , the free electrons that are present catalyze h@xmath1 formation . in turn , h@xmath1 cooling becomes efficient in the center of the halo , and dcbh formation is prevented . we emphasize , however , that in many cases the gas collapsing into atomic cooling haloes at high redshift is self - shielding to ionizing radiation . therefore , it is only in a fraction of such haloes in which dcbh formation is prevented due to reionization . -0.8 cm black holes molecules early universe cosmology : theory
gamma - ray bursts ( grbs ) are instantaneously the most luminous objects in the universe , produced by the deceleration of ultra - relativistic outflow ( lorentz factors @xmath14 ) . the core - collapse of massive stars are the progenitor of long grbs , and the merger of binary compact stellar objects such as neutron stars ( ns ) and black holes ( bh ) are the possible progenitor of short grbs ( woosley & bloom 2006 , nakar 2007 , berger 2014 ) . in both cases accretion onto a compact object is likely to power the relativistic outflow and the same physical processes are involved . the outflow energy is first dissipated by internal shocks ( or another form of internal dissipation ) which produces the prompt @xmath15-rays . later the interaction of the outflow with the ambient medium produces an external shock which expands and produces the subsequent afterglow ( e.g. piran 2004 ; zhang & mszros 2004 ) . relativistic motion is an essential ingredient in the grb model although the exact outflow formation process is not known . understanding the nature of the outflow , especially the acceleration , collimation , and energy content is a major focus of international research efforts in the context of grb and other astrophysical jets . grb outflows are conventionally assumed to be a baryonic jet ( paczynski 1986 ; shemi & piran 1990 ) , although polarization measurements imply that magnetic fields play a role in the jet acceleration ( e.g. steele et al . 2009 ; mundell et al . 2013 ; gtz et al . 2009 ; yonetoku et al . relativistic outflows and possibly magnetic acceleration are features that grbs , active galactic nuclei ( agn ) , and microquasars have in common . stellar tidal disruption by a massive bh is also likely to produce a relativistic jet ( bloom et al . 2011 ; burrows et al . 2011 , zauderer et al . 2011 , levan et al . 2011 , cenko et al . 2012 ) . by studying and comparing the properties of these objects , we could gain an insight into the processes that govern the formation of relativistic jets ( e.g. marscher 2006a , nemmen et al . 2012 ) . in the case of blazars , we can measure apparent superluminal motion ( i.e. lower limits of @xmath1 ) , where reported apparent velocities are as high as @xmath16 for @xmath15-ray bright blazars ( jorstad et al . 2005 ; lister et al 2009 ; piner et al . 2012 , liodakis & pavlidou 2015 ) . the lorentz factor for agn is typically @xmath17 ( e.g. marscher 2006a ; saikia et al . 2016 , etc . ) or @xmath18 ( lister et al . 2009 ) . blazars with a high@xmath19 overpopulate centimeter - wave surveys of bright flat - spectrum sources because of beaming bias . alternatively , a volume - limited sample of radio - loud agn would be dominated by objects with more mundane jets . a power - law distribution of lorentz factors for agn can be assumed , @xmath20 , where population synthesis studies show that a value of @xmath21 between 1.5 and 1.75 provides a good match between a synthetic and observed distribution of apparent velocities ( lister & marscher 1997 ; marscher 2006b ) . recent work indicates a value of @xmath22 for blazars ( saikia et al . 2016 ) . many observations indicate that grbs are produced by ultra - relativistic outflows with @xmath23 . however , grb progenitors might not always eject such a high@xmath19 flow . for example , if the outflow is baryonic , the baryon loading might not always be optimal , resulting in lower lorentz factors . for an outflow with low @xmath1 , the internal dissipation processes ( i.e. @xmath15-ray production ) happen when the outflow is still optically thick . since we are currently discovering grb events through wide field monitoring of the @xmath15-ray sky ( e.g. swift , fermi , ipn ) , a population of low@xmath19 outflows might be undiscovered . compact stellar mergers are the most promising targets for ground - based gravitational wave ( gw ) detectors such as advanced ligo , virgo and kagra . the merger of a binary bh system produced the advanced ligo detection gw150914 , the first direct observation of gw ( abbott et al . em counterparts to bh - bh mergers are not expected , fermi gamma - ray burst monitor ( gbm ) however , claimed a 2.9@xmath24 detection of a weak @xmath15-ray burst 0.4 seconds after the gw detection ( connaughton et al . 2016 ) , if this burst is associated with gw150914 then an electromagnetic ( em ) afterglow would also be present ( yamazaki et al . 2016 ) . to maximize the science returns from further gw detections , the identification of an em counterpart will be crucial . the @xmath15-ray emission from short grbs are an ideal em counterpart to ns - ns / ns - bh mergers , and potentially bh - bh mergers . however , they occur relatively rarely within the range of gw detectors ( 300 mpc for face - on ns - ns mergers ) , this is possibly because @xmath15-ray emission is highly collimated , or the mis - match between short grb peak energies and the swift detection band can make detection more difficult . additionally the intrinsic rate of compact object mergers within this volume is relatively low . more isotropic em components such as macronovae are often discussed to localize a large sample of gw events ( e.g. metzger & berger 2012 ; nakar & piran 2011 ; gao et al . 2013 ; kisaka , ioka & takami 2015 ) . in this paper , we discuss the possibility that a significant fraction of compact stellar mergers result in the production of low@xmath19 jets ( @xmath25 ) . if such jets are common , x - ray , optical , and radio transients , i.e. on - axis orphan afterglows ( dermer et al . 2000 , nakar & piran 2002a , huang et al . 2002 , rhoads 2003 , cenko et al . 2013 , cenko et al . 2015 ) , would be more frequent than short grbs . such low frequency transients would accompany a good fraction of gw events and they allow for the accurate determination of the sky positions of the gw sources . the time lag between gw signals , where we can assume that the jet launch time @xmath26 is coincident with the merging time when the gw amplitude becomes maximal , and em jet emission will enable us to determine the @xmath1 distribution of jets from compact stellar mergers and it will provide another constraint on the acceleration process of relativistic jets . in @xmath27 [ motion ] we discuss the background of relativistic motion in the standard grb fireball model and the implications for the prompt @xmath15-ray emission . @xmath27 [ oaoag ] the case for a population of low lorentz factor jets is made . @xmath27 [ model ] details the assumptions and conditions made by the monte carlo model plus the numerical results . @xmath27 [ gwimp ] highlights the implications for gw rates within the ligo / virgo detection volume . in @xmath27 [ conc ] conclusions are given . observed grbs contain a large fraction of high energy @xmath15-ray photons , which can produce electron - positron pairs if they interact with lower energy photons . if the optical depth for this process is large , pairs will form rapidly and compton scatter other photons , resulting in an increased optical depth . the optical depth for the pair creation is very sensitive to the lorentz factor of the source @xmath28 ( e.g. piran 1999 ; lithwick & sari 2001 for the typical high energy spectral index @xmath29 ) . the source becomes optically thin if it is expanding with a lorentz factor @xmath23 . if there are baryons in grb outflows , another limit on @xmath1 can be obtained by considering the scattering of photons by electrons associated with these baryons ( e.g. lithwick & sari 2001 ) . note that high polarization results still suggest magnetized baryonic jets , rather than poynting - flux dominated jets ( steele et al . 2009 ; mundell et al . the optical depth due to these electrons at radius @xmath30 is @xmath31 where @xmath32 is the thomson cross - section , @xmath33 is the total isotropic explosion energy and @xmath34 proton mass . outflows become optically thin at the photospheric radius , @xmath35 where @xmath36ergs and @xmath37 . on the other hand , the variability timescale @xmath38 in grbs constrains the radius from which the radiation is emitted , @xmath39 where @xmath40 seconds . requiring @xmath41 , we obtain @xmath42 . for outflows with a small lorentz factor @xmath43 , the internal dissipation happens when the outflow is still optically thick . the photons will remain trapped and the thermal energy will be converted back to the kinetic form ( kobayashi & sari 2001 ; kobayashi et al . 2002 ) , and the prompt @xmath15-ray emission would be suppressed ( i.e. failed grbs ) . usually outflows are assumed to have a sub - relativistic temperature after the internal dissipation , and the internal energy density is comparable to the mass energy density @xmath44 . if a significant fraction of the internal energy is converted to electron - positron pairs , the number density of the electrons and positrons @xmath45 could be larger by a factor of @xmath46 than that of electrons that accompany baryons , where @xmath47 is the electron mass . a more detailed discussion ( lithwick & sari 2001 ) also shows that the scattering of photons by pair - created electrons and positrons is nearly always more important than that by electrons that accompany baryons . since the lepton pairs create an effective photosphere further out than the baryonic one , the approximation in equation ( [ eq : rph ] ) will provide conservative estimates when we discuss failed grb rates in @xmath27 4 and @xmath27 5 . even if a jet does not have a velocity high enough to emit @xmath15-rays , it eventually collides with the ambient medium to emit at lower frequencies . such synchrotron shock radiation has been well studied in the context of grb afterglows ( e.g. mszros & rees 1992 , 1997 ; sari & piran 1999 ; kobayashi et al . 1999 ) . because of relativistic beaming , the radiation from a jet can be described by a spherical model when @xmath48 where @xmath49 is the jet half - opening angle . we here consider a relativistic shell with an energy @xmath33 and an initial lorentz factor @xmath1 expanding into ism with particle density @xmath50 . the deceleration of the shell happens at , @xmath51 where @xmath52 protons @xmath53 , and @xmath54 is measured in the grb rest frame . the typical frequency and the spectral peak flux of the forward shock emission at the deceleration time @xmath54 are , @xmath55 ( sari et al . 1998 ; granot & sari 2002 ) where @xmath56 and @xmath57 are the microscopic parameters , @xmath58 , @xmath59 , and @xmath60 cm ( i.e. the ligo range for face - on ns - ns mergers ) . the optical emission , assumed to be between the peak frequency @xmath61 and the cooling frequency @xmath62 , is expected to rise as @xmath63 and decay as @xmath64 after the peak @xmath65 . self - absorption can significantly reduce synchrotron shock emission at low frequencies . the upper limit can be approximated as black body flux for the forward shock temperature ( e.g. sari & piran 1999 ) , the limit at @xmath54 is @xmath66 where @xmath67ghz and the observable blast - wave size @xmath68 . equalizing the synchrotron emission and the black body limit , we obtain the self - absorption frequency @xmath69 ghz at the deceleration time @xmath54 . the self - absorption limit initially increases as @xmath70 , and then steepens as @xmath71 after @xmath61 crosses the observational frequency @xmath72 . considering that the synchrotron flux at @xmath73 also increases as @xmath70 , if @xmath74 at @xmath54 , the synchrotron emission would be reduced by the self - absorption at least until the passage of @xmath61 through the observational band at @xmath75 days . if the jet break happens while the flux is still self - absorbed , the light curve becomes flat @xmath76 const ( sari , piran & halpern 1999 ) . however , this estimate is obtained by assuming the rapid lateral expansion ( i.e. @xmath77 ) . recently studies show that the sideways expansion is rather slow especially for mildly - relativistic jets ( granot & piran 2012 ; van eerten & macfadyen 2012 ) . we will assume that the blast - wave emission starts to decay at the jet break , @xmath78 even if it is in the self - absorption phase . at low frequencies @xmath79 ghz and early times , forward shock emission would be affected by synchrotron self - absorption . however , currently most radio afterglow observations are carried out at higher frequencies ( e.g. vla 8.5 ghz ) at which self - absorption is more important for the reverse shock emission . just before the deceleration time @xmath54 , a reverse shock propagates through the jet and heats the original ejecta from the central engine . the reverse shock region contains energy comparable to that in the forward shock region . however , it has a lower temperature due to a higher mass ( i.e. lower energy per particle ) . the shock temperature and the typical frequency are lower by a factor of @xmath80 and @xmath81 compared to those of the forward shock ( e.g. kobayashi & zhang 2003 ) . although reverse shocks in low-@xmath1 jets could emit photons in the radio band , the self - absorption limit is tighter due to the lower shock temperature ; we find that the forward shock emission always dominates . note that we rarely catch the reverse shock emission even for regular grbs with detectable @xmath15-ray emission . we will discuss only the forward shock ( i.e. blast wave ) emission in this paper . by using the estimates of lorentz factors based on long grb afterglow peak times , hascot et al . ( 2014 ) demonstrated that an apparent correlation between isotropic @xmath15-ray luminosity @xmath82 and lorentz factor @xmath1 can be explained by a lack of bright bursts with low lorentz factors . they have also predicted the existence of on - axis orphan afterglows of long grb events . we here extend their argument to short grbs , and we apply their formalism to cosmological ( i.e. @xmath15-ray satellite range ) and local ( i.e. gw detector range ) events to study the on - axis orphan afterglows of failed short grbs ( i.e. low-@xmath1 events ) . the following assumptions are made in our simple monte carlo simulation of a synthetic population of merger events : 1 . the redshift for each event is randomly determined using a distribution with a constant time delay with respect to the star formation rate , where the peak rate is at @xmath83 . the redshift limits of @xmath84 are used for the cosmological sample , and @xmath85 for local sample , i.e advanced ligo / virgo detectable range @xmath86 mpc = 300 mpc for ns - ns mergers where the factor of 1.5 accounts for the stronger gw signal from face - on mergers ( kochanek & piran 1993 ) . we use the event rate per unit comoving volume for short grbs obtained by wanderman & piran ( 2015 ) , which is a function of @xmath87 as @xmath88 numerical results for the cosmological cases are insensitive to the value of the maximum @xmath87 as long as it is much larger than unity . 2 . a power - law distribution of lorentz factors @xmath89 is assumed with reasonable limits @xmath90 . motivated by agn studies ( e.g. lister & marscher 1997 ; marscher 2006b ) , we choose @xmath91 as our fiducial value and the cases of @xmath92 and @xmath93 will be briefly discussed . the isotropic @xmath15-ray luminosity @xmath82 is randomly generated in the limit @xmath94erg / s @xmath95erg / s where the limits come from observational constraints and the luminosity distribution follows the form obtained by wanderman & piran ( 2015 ) , @xmath96 where this luminosity function is logarithmic in the interval d@xmath97 . for each event , the dissipation radius @xmath98 is evaluated by using a random @xmath1 and the typical pulse width in short grb light curves @xmath99 sec ( nakar & piran 2002b ) . @xmath15-ray photons are assumed to be emitted at @xmath100 with a random @xmath15-ray luminosity @xmath101 or equivalently a random isotropic @xmath15-ray energy @xmath102 where @xmath103 is the duration of short grbs . we assume @xmath104 sec for all bursts as this is the median value for a log normal distribution of durations for short gamma - ray bursts ( zhang et al . the spectral peak energy in the @xmath105 spectrum is known to be correlated with @xmath101 ( yonetoku et al . 2004 , ghirlanda et al . the correlation is consistent for both long and short grbs ( zhang et al . 2012 ) , and given by @xmath106 the @xmath105 spectrum is assumed to follow a broken power - law with low - energy index ( below @xmath107 ) of @xmath108 ( @xmath109 ) , and a high - energy index of @xmath110 ( @xmath111 ) , where @xmath112 and @xmath113 are the photon number spectral indices . the mean index values for all grbs are @xmath114 and @xmath115 ( gruber et al . 2014 ) but as short grbs are typically harder than average we use the values @xmath116 and @xmath117 . the spectral peak is normalized as the value integrated between 1 kev and 10 mev giving @xmath101 . if the outflow is optically thin , all the photons released at @xmath118 are radiated away . the event is considered to be detectable if the photon flux at the detector in the swift band ( 15 - 150 kev ) is @xmath119 photons s@xmath120 @xmath121 ( band 2006 ) . we take into account the redshift of the spectrum when the photon flux is evaluated . if the optical depth at the dissipation radius @xmath100 is more than unity , or equivalently the photospheric radius @xmath122 is larger than the dissipation radius , the @xmath15-ray emission would be suppressed where @xmath123 is the explosion energy and @xmath124 is the conversion efficiency from the explosion energy to @xmath15-rays . we use @xmath125 , this is consistent with theoretical predictions ( kobayashi et al . 1997 ) and the fiducial value used in other works ( liang et al . 2010 ; ghirlanda et al . the @xmath15-ray energy injected at @xmath100 is adiabatically cooled , and the photons decouple from the plasma at @xmath126 . assuming a sharp transition from optically thick to thin regime ( see beloborodov 2011 for the discussion of fuzzy photosphere ) , we use hydrodynamic scalings to estimate the cooling factor . the internal energy density ( photon energy density ) decays as @xmath127 and the lorentz factor is constant for the outflow with a sub - relativistic temperature ( piran et al . considering that the internal energy in the outflow shell with width @xmath128 is @xmath129 , the luminosity of photons released at @xmath118 is @xmath130 where we have assumed no shell spreading @xmath131 const . the spectral peak energy is similarly shifted as @xmath132 . the photons in the coupled plasma undergo pair production and compton down - scattering that progressively thermalises the distribution ( hascot et al . the electron temperature at @xmath118 can be approximated by a black - body temperature @xmath133 where @xmath21 is the radiation constant . the optical depth at @xmath118 is given by @xmath134 . the condition for efficient thermalisation is @xmath135 ( peer et al . 2005 , thomson et al . 2007 ) where @xmath47 is the mass of an electron and @xmath136 the boltzmann constant . the peak energy @xmath107 for such a case is given by @xmath137 , above which the distribution is exponentially suppressed . for simplicity we assume @xmath138 . if @xmath139 , the photons are not efficiently thermalised . the distribution is then limited by the efficiency of pair production where the maximum energy is @xmath140 kev . the distribution is cut - off above this energy . we generate a sample of @xmath141 events and evaluate the @xmath15-ray flux for each in the swift band . to allow for clarity without losing the general trend , the results for a population of 2000 events are shown in figure [ fig : f1 ] ; the blue circles and red crosses show the events detectable and undetectable by swift , respectively . the isotropic kinetic energy @xmath142 is the energy in the blast wave after deceleration time , @xmath143 , where @xmath33 is the total isotropic explosion energy , and @xmath144 is the isotropic @xmath15-ray energy at the photospheric radius @xmath126 . the lorentz factor @xmath1 of an outflow at @xmath145 is shown against this . the top panel shows the results with @xmath84 , where we find a small fraction @xmath146 of the total population and @xmath147 of the events with @xmath148 are detectable by swift . for the local population @xmath149 , these fractions are higher , at @xmath150 and @xmath151 respectively , due to the proximity ( see the bottom panel ) . the dashed line indicates the lower limit for a successful grb , events below this line have the prompt @xmath15-ray emission fully suppressed ; the cut - off , with the parameters used , is given by @xmath152 . in figure [ fig : f1 ] , the low - energy limit of @xmath153 is basically set by the monte - carlo luminosity distribution ( i.e. @xmath154 erg / s . note that the explosion energy @xmath33 is higher than the @xmath15-ray energy @xmath155 at the dissipation radius @xmath100 by a factor of @xmath156 ) . if we consider the local population ( the bottom panel ) , for the events above the dashed line ( i.e. the blue circles ) all of the @xmath15-ray energy is successfully radiated away , whereas for the events below the dashed line ( i.e. the red crosses ) , almost all of the @xmath15-ray energy is reabsorbed into the outflow . thus the distribution of @xmath153 for the blue circles has a slightly lower limit . if we consider the cosmological population ( the top panel ) , a fraction of events are distant and intrinsically dim . they are undetectable by swift even if all gamma - ray energy is successfully radiated away at @xmath100 . this is why there are red crosses above the dashed line for the cosmological population . the fraction of the events detectable by swift weakly depends on @xmath157 . if we assume @xmath158 erg / s , swift would be able to detect @xmath159 of the total cosmological population , and @xmath160 of the total local population . liang et al . ( 2010 ) , ghirlanda et al . ( 2012 ) and tang et al . ( 2015 ) report correlations between lorentz factor @xmath1 and the isotropic luminosity @xmath101 ( or the isotropic energy @xmath161 ) for long grbs : @xmath162 ; @xmath163 ; and @xmath164 , respectively . however , such power - law relations could indicate a lower limit on @xmath1 for observable long grbs with a given burst energy ( hascot et al . 2014 ) . in our simulation , we find that the detectable short bursts are always located above a line @xmath165 giving a lower limit relation @xmath166 . as discussed in section 3 , the kinetic energy @xmath153 of the failed grbs will be released as on - axis orphan afterglows at late times . figure [ fig : f2 ] shows the distributions of the peak flux ( the top panel ) and peak time ( the bottom panel ) of such x - ray , optical , and radio transients . to estimate these distributions , we have used the monte carlo results for the local sample ( @xmath167 mpc ) with model parameters : @xmath168 protons @xmath53 ( berger 2014 ; metzger & berger 2012 ) , @xmath169 , @xmath170 ( panaitescu & kumar 2002 ; yost et al . 2003 ; berger 2014 ) , the index of the power - law distribution of random electrons accelerated at shock @xmath171 ( sari , narayan & piran 1996 ; daigne et al 2011 ; metzger & berger 2012 ) , and the jet half - opening angle @xmath172 ensuring @xmath173 for our sample and is within the limits @xmath174 found by fong et al . ( 2015 ) for short grb . the jet opening angle plays a role only when we estimate the jet break time . the dotted green lines ( figure [ fig : f2 ] ) indicate the distribution for x - ray transients . the typical frequency of the blast wave emission @xmath61 is sensitive to the lorentz factor @xmath175 . since for the local population the on - axis orphan afterglows are produced by low-@xmath1 jets ( @xmath176 ) , the typical frequency @xmath61 is expected to already be below the x - ray and optical band at the deceleration time @xmath54 . the x - ray and optical light curves should peak at @xmath54 and they have the same peak time distribution . considering that the deceleration time @xmath177 is mainly determined by @xmath1 , we can roughly estimate the peak - time distribution @xmath178 . for @xmath179 , the distribution is wide and a large fraction of the events have the peak - time @xmath54 around several days after the merger event . if the minimum lorentz factor @xmath180 is assumed , the peak - time distribution would achieve the peak around a few weeks after the merger event . the distribution of the peak flux for x - ray , where the frequency is above the cooling frequency @xmath181 , is @xmath182 is shown in the top panel . given good localisation , all of the x - ray peak afterglow flux is above the minimum senstivity of the swift xrt @xmath183 erg @xmath121 s@xmath120 for @xmath184 seconds ( the vertical green thick solid line ) . the x - ray afterglows are below the trigger sensitivities of swift bat and maxi ; and too faint to be detectable by the swift bat survey . the solid red line in the top panel and the dotted green line in the bottom panel indicates the distribution for optical ( _ g_-band ) transients . the ab magnitude m@xmath185 axis is added in the top panel to indicate the optical flux . for optical transients , peak flux is @xmath186 , and @xmath187 of the optical orphan afterglows are brighter than @xmath188 ( the vertical solid red line indicates this typical limit for mid - sized ( @xmath189 m ) telescopes ) . the peak - time distribution for the bright events ( @xmath190 ) is shown as the the dashed magenta line in the bottom panel . the difference between the dotted green ( representing both x - ray and optical in peak time ) and dashed magenta line corresponds to the dim event population ( @xmath191 ) . since these events tend to have low-@xmath1 , their typical frequencies are much lower than the optical band , and they peak at late times . the solid blue lines give the distribution for radio ( 10 ghz ) transients . the typical frequency @xmath61 is expected to be above 10 ghz at the deceleration time @xmath54 . the light curve peaks when the typical frequency @xmath192 crosses the observational band : @xmath193 . since the dynamics of the blast wave at @xmath194 depends only on the sedov length @xmath195 and not on the initial lorentz factor @xmath1 , the peak - time distribution should be narrowly clustered , compared to the distribution of the optical transients . the monte carlo results actually give a narrow peak around @xmath196 10 days . the peak flux @xmath197 is bright : typically @xmath198 mjy . vla ( the vertical solid blue line ) can easily detect the transients . the dashed - dotted black lines indicate the distribution for radio ( 150 mhz ) transients . as we have discussed , this low frequency emission is suppressed by the self - absorption , and jet break is likely to happen before it becomes optically thin . the peak - time of the light curve is determined by the jet break time @xmath199 . for the fixed @xmath172 , we find that the peak - time distribution is similar to that for 10 ghz transients and it peaks around @xmath200 days . however , since the emission is still suppressed by the self - absorption at the peak time , the peak flux is much lower : @xmath201 mjy . approximately @xmath202 of the 150 mhz transients are brighter than the sensitivity limit of 48 lofar stations ( the vertical dashed black line ) , and all are brighter than the sensitivity limit for ska1-low ( the vertical dashed - dotted black line ) . typical afterglow light curves for a selection of on - axis orphan afterglows are shown in figure [ fig : f4 ] . an average luminosity distance for ns - ns gw detectable mergers from our sample is used of @xmath203 220 mpc . x - ray , optical , and radio ( 10 ghz ) are shown for 4 combinations of @xmath1 and @xmath153 . the vertical dashed line in each panel represents the deceleration time @xmath54 , as @xmath54 is most sensitive to @xmath1 ( see equation [ tdec ] ) the lower lorentz factor cases ( top two panels ) have a significantly later deceleration time . the vertical dotted line in each panel represents the jet - break time @xmath204 , a jet half - opening angle @xmath172 is used throughout , for narrower(wider ) jet half - opening angles the break time will be at earlier(later ) times . the jet - break time is only weakly dependent on the kinetic energy ( see equation [ eq : tb ] ) . in all cases the x - ray ( green dash - dotted line ) and the optical ( thin red line ) peak at the deceleration time , the 10 ghz ( thick blue line ) is shown to peak at a later time @xmath205 when the typical frequency @xmath206 crosses the radio frequency . in all cases at times earlier than @xmath54 the flux is @xmath207 , for the x - ray and optical the flux at @xmath208 is @xmath209 . at 10 ghz the flux is @xmath210 at @xmath211 , and @xmath212 after @xmath205 and before @xmath204 . in all cases at @xmath213 the flux is @xmath214 . the swift satellite has been detecting short grbs at a rate of @xmath215 yr@xmath120 since the launch in 2004 , and @xmath216 of the detected events have measured redshifts ( swift grb catalogue ) . unfortunately no swift short grb with known redshift has been detected within the advanced ligo / virgo range for face - on ns - ns mergers @xmath217 mpc , and only three ( 061201 , 080905a , and 150101b ) have occurred within the face - on ns - bh range @xmath218 mpc ( abadie et al . metzger & berger ( 2012 ) estimate that @xmath219 short grbs per year , with redshift measurements , are currently being localized by swift within @xmath220mpc ( @xmath221mpc ) . considering that the field of view of the swift bat is @xmath189sr , the all - sky rate of detectable short grbs with or without redshift information is higher by a factor of @xmath222 . if the distribution of @xmath1 is described by the power - law @xmath89 , when we consider the rate of jets from mergers regardless of inclination or detectability , the rate for failed grbs would be higher than the short grb rate . for local population @xmath223 mpc , we find that the fraction of failed events is about @xmath224 for @xmath225 , @xmath226 for @xmath227 , and @xmath228 for @xmath93 ( the same rates are obtained for a population of @xmath229mpc ) . if @xmath230(@xmath93 ) , the failed grb rate is higher by a factor of @xmath231(6.7 ) than the short grb rate ( i.e. the ratio of failed to successful grbs ) . the all - sky rate of the failed grbs with or without redshift information is about 2.6(5.1 ) per year for the ns - ns range and 26(51 ) per year for the ns - bh range . here we assumed the jet opening angle distribution does not depend on the lorentz factor of the jets ( i.e. grb and failed grb jets have the same opening angle ) . the jet half - opening angle is not well constrained for short grb jets ( the median value for 248 long grbs is @xmath232 ; fong et al . 2015 ) . using four short grbs which have temporal steepenings on timescale of @xmath233 days , the median value is estimated as @xmath234 ( fong et al . however , the majority of short grbs do not have detected jet break , the inclusion of these bursts is essential in understanding the true opening angle distribution . based on a probability argument , fong et al . ( 2015 ) obtain the median value @xmath235 16@xmath236 and 33@xmath236 if the maximum possible angle is 30@xmath236 and 90@xmath236 , respectively . if the typical jet half - opening angle of short grbs is @xmath237 , the beaming factor is @xmath238 where @xmath239 , only a small fraction of short grb jets point toward us ( see the black dashed line in figure [ fig : f3 ] ) . however , since the gw polarization components @xmath240 and @xmath241 depend on the inclination angle @xmath242 of the binary , mergers emit gws much more strongly along the polar axis than in the orbital plane . considering that the jets from the mergers are also likely to be directed along the polar axis , kochanek & piran ( 1993 ) show that when a grb is associated , the gw amplitude @xmath243 is stronger by a factor of @xmath244 than the amplitude averaged over the sky ( as seen from the source ) . the distances out to which gw detectors could detect the binary increases by a factor of @xmath245 if the jet points toward us ( we define an on - axis event as any jet where the inclination is within the half - opening angle , @xmath246 ) . when we consider a sample of merger gw events detected by a gw detector with sensitivity @xmath247 , their jets would tend to be directed to us . this is because on - axis events are detectable at a larger distance . the on - axis probability could be higher by roughly the volume factor of @xmath248 ( the blue dashed - dotted line , figure [ fig : f3 ] ) than the simple geometric estimate @xmath249 ( i.e. our line - of - sight falls within the opening angle of the jet with a higher probability ) . we also conduct a monte carlo simulation to estimate the on - axis probability . in the simulation , mergers are uniformly distributed in space , with a random inclination angle , and they emit gws with amplitude @xmath250 . after selecting the events detectable by a gw detector : @xmath251 , we evaluate the fraction of the events which have an inclination angle smaller than a given jet half - opening angle @xmath49 ; we assume uniform jets with a top hat distribution throughout from the jet symmetry axis ( e.g. @xmath252 outside of some core angle ) , only the central part could have lorentz factors high enough to produce @xmath15-rays . although the detailed study is beyond the scope of this paper , the failed grb rate could be even higher for structured jets . the result ( the red solid line ) does not depend on the detector sensitivity as long as the merger distribution is homogeneous . if we consider gw trigger events , the on - axis probability ( the red solid line ; @xmath253 and @xmath254 for @xmath255 and 33@xmath236 , respectively ) is much higher than the beaming factor ( the black dashed line ) . although isotropic em counterparts such as macronovae could be ideal to localize a large sample of gw events , @xmath256 of gw events would still be associated with the on - axis orphan afterglow of failed grbs especially when they have wider jet opening angles compared to short grb jets . for long grb jets , observational results indicate such a correlation @xmath257 with @xmath258 ( panaitescu & kumar 2002 ; salmonson & galama 2002 ; kobayashi et al . 2002 ; ghirlanda et al . the failed grb rates could be higher than those discussed at the beginning of this section . we have shown that failed grbs are much more frequent than short grbs when the lorentz factors of jets from compact stellar mergers follow a similar power - law distribution as those observed for agn . for most events the internal dissipation process happens when the jet is still optically thick , and the photons produced by the dissipation process will be converted back to the kinetic energy of the jet . by using a simple monte carlo model , we have shown that even for the local merger population within the ligo / virgo range , the @xmath15-ray emission from jets with @xmath176 will not be detected by @xmath15-ray satellites ( e.g. swift ) . for a power - law distribution of the jet lorentz factors @xmath259 , 78% of compact object mergers that have jets result in a failed grb . the failed grb events will produce on - axis orphan afterglow at late times . using the local short grb rate as normalization , the all - sky rate of the on - axis orphan afterglow is about 2.6 and 26 per year for the ns - ns range ( 300 mpc ) and ns - bh range ( 600 mpc ) , respectively . the opening angle of jets for long grbs was found to be a function of @xmath1 ( e.g. ghirlanda et al . 2013 ) , if low-@xmath1 jets from compact - binary mergers have wider half - opening angles @xmath49 than those of short grbs then the real rate would be higher than these . we have evaluated the peak time and peak luminosity of the on - axis orphan afterglows in x - ray , optical , and radio bands . although it is usually difficult to model observational data for orphan afterglow candidates when the explosion time is unknown ( i.e. the @xmath26 issue ) . for gw trigger events , gw signals will provide the explosion time @xmath26 . the peak time distribution in the x - ray and optical band is rather wide @xmath260 days after the gw signals . although the sky localization of sources by gw detectors is not accurate enough for follow - up observations by most conventional telescopes ( abbott et al . 2016a ) , 85% of the on - axis orphan afterglows are brighter than @xmath261 . the current and upcoming optical transient search ( e.g. iptf / ztf , pan - starrs , goto , blackgem , kiso , skymapper , subaru hsc , lsst ) should be able to detect the optical transients . the x - ray and/or optical detection can be followed by radio observations ( e.g. vla ) , also several radio instruments have the potential to be leading transient detectors due to their large fov ( e.g. ska , lofar , apertif , mwa ) . radio emission is expected to peak around @xmath262 days after the merger events . optical and radio observations will constrain the opening angle of low-@xmath1 jets ( and high @xmath1-jets ) . since merger jets from gw trigger events tend to be directed to us , the on - axis probability ( e.g. 13% and 44% for @xmath255 and 33@xmath236 , respectively ) is much higher than the beaming factor @xmath263 . a significant fraction of gw events could be associated with on - axis orphan afterglows . observations of on - axis orphan afterglows and grb afterglows will enable us to determine the @xmath1 distribution of jets ( e.g. clustered at high-@xmath1 , a power - law distribution , a lognormal , or multiple populations ) , and it will provide constraints on the acceleration process of relativistic jets . abadie , j. et al . 2010 , class . quantum grav . , 27 , 173001 . abbott , b. p. et al . 2016a , living reviews in relativity , 19 , 1 abbott , b. p. et al . 2016b , phys . lett . , 116 , 06112 . band , d. l. 2006 , apj , 644 , 378 . beloborodov , a.m. 2011 , apj , 737 , 68 . berger , e. 2014 , annu . , 52 , 43 . bloom , j.s . 2011 , science , 333 , 203 . burrows , d.n . 2011 , nature , 476 , 421 . cenko , s.b . 2012 , apj , 753 , 77 . cenko , s.b . 2015 , apj , 803 , 24 . cenko , s.b . 2013 , apj , 769 , 130 . connaughton , v. et al . 2016 , arxiv : 1602.03920 . daigne , f. et al . 2011 , a&a , 526 , 13 . dermer , c. d. et al . 2000 , apj , 537 , 785 - 795 . fong , w. et al . 2015 , arxiv:1509.02922 gao , h et al . 2013 , apj , 771 , 86 . ghirlanda , g. et al . 2009 , a&a , 496 , 585 . ghirlanda , g. et al . 2012 , mnras , 420 , 483 . ghirlanda , g. et al . 2013 , mnras , 428 , 1410 - 1423 . granot , j. & sari , r. 2002 , apj , 568 , 820 . granot , j. & piran , t. 2012 , mnras , 421 , 570 . gruber , d. et al . 2014 , apjs , 211 , 27 . gtz , d. et al . 2009 , apjl , 695 , l208 . hascot , r. et al . 2014 , apj , 782 , 5 . huang , y. f. et al . 2002 , mnras , 332 , 735 . jorstad , s.g . 2005 , aj , 130 , 1418 . kisaka , s. , ioka , k. & takami , h . 2015 , apj , 802 , 119 . kobayashi , s. piran , t. & sari , r. 1997 , apj , 490 , 92 . kobayashi , s. , piran , t. , & sari , r . 1999 , apj 513 , 669 . kobayashi , s. & sari , r. 2001 , apj 551 , 934 . kobayashi , s. , ryde , f. & macfadyen , a . 2002 , apj 577 , 302 . kobayashi , s. & zhang , b. 2003 , apj 582 , l75 . kochanek , c & piran , t. 1993 , apj , 417 , l17 . levan , a.j . 2011 , sci , 333 , 199 . liang , e. et al . 2010 , apj , 725 , 2209 . lister , m. et al . 2009 , apj , 138 , 1874 . lister , m. & marscher , a.p . 1997 , apj , 476 , 572 . lithwick , y. & sari , r. 2001 , apj , 555 , 540 . liodakis , i. & pavlidou , v. 2015 , mnras , 451 , 2434 . marscher , a.p . 2006a , in relativistic jets : the common physics of agn , microquasars , and gamma - ray bursts , ed . hughes & j.n . bregman , aip conf . marscher , a.p . 2006b , pos , proceedings of the vi microquasar workshop : microquasars and beyond . metzger , b. d. & berger , e. 2012 , apj , 746 , 48 . mszros , p. & rees , m. 1992 , mnras 258 , 41 . mszros , p. & rees , m. 1997 , apj 476 , 232 . mundell , c. et al . 2013 , nature , 504 , 119 . nakar , e. & piran , t. 2002a , new ast . , 8 , 141 - 153 . nakar , e. & piran , t. 2002b , mnras , 330 , 920 . nakar , e. 2007 phys . , 442 , 166 - 236 . nakar , e. & piran , t. 2011 , nature , 478 , 82 . nemmen , r. s. 2012 , science , 338 , 1445 . paczyski , b. 1986 , apj , 308 , l43 . panaitescu , a . & kumar , p. 2002 , apj , 571 , 779 . peer , a. et al . 2005 , apj , 635 , 476 - 480 . piner , b.g . 2012 , apj , 758 , 84 . piran , t. et al . 1993 , mnras , 263 , 861 - 867 . piran , t. 1999 , phys . rept . , 314 , 575 piran , t. 2004 , rev . mod . phys . , 76 , 1143 . rhoads , j. e. 2003 , apj , 591 , 1097 . saikia , p. , elmar , k. & falcke , h. 2016 , mnras , 978 . salmonson , j.d . & galama , t.j . 2002 , apj , 569 , 682 . sari , r. , narayan , r. & piran , t. 1996 , apj , 473 , 204 . sari , r. & piran , t. 1999 , apj 520 , 641 . sari , r. , piran , t. & halpern , j.p . 1999 , apjl 519 , l17 . sari , r. , piran , t. & narayan , r. 1998 , apjl 497 , l17 . shemi , a. & piran , t. 1990 , apj 365 , l55 . steele , i et al . 2009 , nature , 462 , 767 . tang , q. w. et al . 2015 , apj , 806 , 194 . thomson , c. 2007 , apj , 666 , 1012 - 1023 . van eerten , h. j. & macfadyen , a. i. 2012 , apj , 751 , 155 . wanderman , d. & piran , t. 2015 , mnras , 448 , 3026 . woosley , s. e. & bloom , j. s. 2006 , annu . astrophys . , 44 , 1 . yamazaki , r. et al . 2016 , arxiv : 1602.05050 . yonetoku , d. et al . 2004 , apj , 609 , 935 . yonetoku , d. et al . 2011 , apjl , 743 , l30 . yost , s. et al . 2003 , apj , 597 , 459 - 473 . zauderer , b. a. et al . 2011 , nature , 476 , 425 - 428 . zhang , b. & mszros , p. 2004 , int . 19 , 2385 . zhang , f. et al . 2012 , apj , 750 , 11 . vs bulk lorentz factor @xmath1 . _ monte carlo _ generated synthetic population of bursts . top panel : cosmological sample of events with @xmath84 . bottom panel : local sample of events with @xmath85 . bursts with prompt emission flux above the swift sensitivity are shown as the blue circles . failed grbs are indicated by the red crosses . @xmath230 is assumed . ] and isotropic kinetic energy @xmath153 . the top(bottom ) two panels have a @xmath265 , and the left(right ) panels have an energy @xmath266 erg . x - ray afterglow are shown as dashed green lines , optical are shown as red thin solid lines , and radio ( 10 ghz ) are shown as blue thick solid lines . the vertical black dotted lines represent the deceleration time @xmath54 and the jet - break time @xmath204 ( assuming a @xmath172 ) ]	short gamma - ray bursts ( grbs ) are believed to be produced by relativistic jets from mergers of neutron - stars ( ns ) or neutron - stars and black - holes ( bh ) . if the lorentz - factors@xmath0of jets from compact - stellar - mergers follow a similar power - law distribution to those observed for other high - energy astrophysical phenomena ( e.g. blazars , agn ) , the population of jets would be dominated by low-@xmath1 outflows . these jets will not produce the prompt gamma - rays , but jet energy will be released as x - ray / optical / radio transients when they collide with the ambient medium . using monte carlo simulations , we study the properties of such transients . approximately@xmath2of merger - jets@xmath3mpc result in failed - grbs if the jet@xmath0follows a power - law distribution of index @xmath4 . x - ray / optical transients from failed - grbs will have broad distributions of their characteristics : light - curves peak @xmath5days after a merger ; flux peaks for x - ray @xmath6mjy ; and optical flux peaks at @xmath7 . x - ray transients are detectable by swift xrt , and @xmath8 of optical transients will be detectable by telescopes with limiting magnitude @xmath9 , for well localized sources on the sky . x - ray / optical transients are followed by radio transients with peak times narrowly clustered around @xmath10days , and peak flux of @xmath11mjy at 10@xmath12ghz and @xmath13mjy at 150@xmath12mhz . by considering the all - sky rate of short grbs within the ligo / virgo range , the rate of on - axis orphan afterglows from failed - grb would be 2.6(26 ) per year for ns - ns(ns - bh ) mergers , respectively . since merger jets from gravitational - wave ( gw ) trigger events tend to be directed to us , a significant fraction of gw events could be associated with the on - axis orphan afterglow .
despite years of study , we still do not know how the seeds of supermassive black holes formed . few if any of the pathways in martin rees s famous flow chart ( begelman & rees 1978 ) can be ruled out , but none of the routes is particularly well understood , either . what we do know is that some very massive ( @xmath2 ) black holes had to exist by @xmath3 in order to explain early quasars ( fan 2006 ) . if the seeds of these black holes were the remnants of massive stars , then they must have grown by eddington - limited accretion for most of the time since their formation , or else much of their growth was due to mergers . a second possibility is that the seeds formed by such a rapid accumulation of matter that it may be considered to be a direct collapse . i will focus on the latter possibility in this paper . the pop iii star formation processes we have heard about at this conference result from the infall of gas at rates @xmath4 yr@xmath5 . what would happen if the infall rate were much higher ? the entropy of matter laid down by gravitational infall onto a growing central mass increases with time . at low inflow rates , however , nuclear ignition halts the contraction of the core and raises the entropy in the interior , leading to a high - entropy object a star . if the inflow rate is high enough , however , the core will be so tightly bound by the time nuclear reactions start that the energy release will be insufficient to halt core contraction . in this case , we are left with an object with a low - entropy core and a high - entropy envelope . this is the situation that can lead to the direct formation of a black hole , without a stellar precursor . this situation should apply when the infall rate exceeds a few tenths of a solar mass per year ( begelman et al . 2006 ) , although more work needs to be done to refine this estimate . the conditions under which such high inflow rates might occur are very uncertain . such rapid infall would almost certainly be driven by gravitational torques , which could be local ( gammie 2001 ) or global ( as in the bars within bars " mechanism : shlosman et al . the natural " gravitational inflow rate is given by @xmath6 yr@xmath5 , where @xmath7 represents the internal velocity dispersion ( @xmath8 turbulent or sound speed ) for a locally unstable thin disk and the orbital speed for a globally unstable ( fully self - gravitating ) system . this means that very large inflow rates are possible in dark matter haloes with velocity dispersions exceeding about 10 km s@xmath5 , which have masses exceeding @xmath9 and become common at redshifts @xmath10 . global gravitational instabilities could occur in a significant fraction of such haloes if the gas is unable to cool much below the virial temperature ( @xmath11 k ) , which requires both that they have been spared significant metal enrichment and that h@xmath12 formation is suppressed ( bromm & loeb 2003 ; begelman et al . however , recent calculations by wise & abel ( 2007 , and these proceedings ) suggest that molecular hydrogen formation is inevitable . this may have the additional effect that many more haloes formed pop iii stars at higher redshifts , making it harder to avoid metal enrichment and possibly depleting the supply of gas available for infall as haloes merge . rapid infall may be easier in more massive haloes . there will be a larger disparity between the thermal temperature of the gas and the virial temperature , but presumably the gas will form a multiphase structure with turbulent velocities ( perhaps driven by stellar energy sources ) dominating the internal energy much like the interstellar medium . simulations are beginning to elucidate these structures , but we have a long way to go before we understand the tradeoffs between large scale inflow and in situ fragmentation and star formation . inflow may be stimulated by the large - scale gravitational torques associated with mergers . a key element in the simulations of black - hole growth and fuelling ( e.g. , di matteo et al . finally , we note that such inflow rates must be possible , because they are required to power quasars . the open question is whether they can also occur when the black hole is absent , or very small . for the remainder of this article we will assume that the answer is affirmative . once the mass of accumulated gas exceeds a few solar masses , radiation pressure dominates the envelope . the core , with gas pressure comparable to radiation pressure , maintains a roughly constant mass @xmath13 . the boundary between dynamic infall and quasistatic contraction is close to the radius where infalling gas liberates energy at the eddington limit . for a constant infall rate of @xmath14 , this radius turns out to be constant , @xmath15 au . the core shrinks under the increasing pressure of the envelope , @xmath16 , where @xmath17 is the envelope mass in solar units . one can sketch the likely interior structure of the envelope using scaling arguments . since the interior is undergoing kelvin helmholtz contraction , the density profile adjusts so that the diffusion timescale @xmath18 is of order the elapsed time at all radii . this means that the specific entropy of any mass shell , @xmath19 , declines slowly ( logarithmically ) with time , and is never far below the value it had when the mass shell was added . in the radiation pressure - dominated envelope , @xmath20 , where @xmath21 and @xmath22 are the radiation and gas pressure , respectively . there are then two possible scaling laws for the envelope structure . if @xmath23 , then @xmath24 , @xmath25 ( for uniform opacity @xmath26 ) , and the envelope joins smoothly onto the core . this was the result presented in begelman et al . the other possible scaling law has @xmath27 , i.e. , nearly isothermal structure . the density profile in the envelope is then @xmath28 , and there is a large jump in density and temperature going from the envelope to the core . the latter structure seems to be the one that emerges naturally from nearly self - similar " models for the core+envelope , which neglect rotation . given the manner in which the mass accumulates , though , rotation is bound to be important . moreover , even a small amount of it can affect the structure dramatically , increasing the binding energy of the gas ( which is otherwise very weakly bound since the mean equation of state has @xmath29 ) and protecting it from pulsational instabilities . including the effects of rotation will require the modeling of angular momentum transport and dissipation , but these complications are not prohibitive . it will be interesting to see which of the scalings applies when rotation is included . fortunately , the behavior of the core is insensitive to the structure of the envelope . since the core is largely dominated by gas pressure , its temperature must track the virial temperature , @xmath30 . rapid nuclear burning starts in the core when the envelope mass is @xmath31 and the density is not that different from the cores of main sequence stars , but the system passes through this phase so quickly ( @xmath32 yr ) that the energy release has little effect . by the time the envelope mass reaches several thousand solar masses , the core temperature has climbed to a few @xmath33 k. because of the extremely deep potential well created by the envelope , the energy released by nuclear reactions at this stage is unable to unbind the gas ( especially if rotation has increased its binding energy ) and the core proceeds to the temperatures ( @xmath34 k ) at which runaway neutrino losses occur . at this point the core loses pressure support and collapses to form a @xmath35 black hole . at the time of its formation , the black hole is embedded in an envelope of more than a hundred times its mass . if it were limited to accreting at the eddington limit for the mass of the black hole , the pressure of the escaping radiation would have essentially no effect on the envelope . in fact , the pressure of the envelope should be able to drive the accretion rate up to the point where the liberated luminosity approaches the eddington limit _ for the mass of the envelope_. relative to the mass of the black hole , the accretion is super - eddington by a factor @xmath36 . accretion inside the massive envelope ( which itself continues to grow at a rate @xmath37 ) can lead to very rapid growth of the black hole . the eddington ratio for black hole growth is @xmath38 where @xmath39 is the accretion efficiency . for an initial black hole mass of @xmath40 , the e - folding time is a thousandth of the salpeter time , or only @xmath11 yr , and super - eddington growth appears possible up to black hole masses @xmath41 . however , this estimate does not take fully into account the back reaction of black hole growth on the envelope . we will see below that the era of rapid growth is limited to much smaller ( but still interesting ! ) black hole masses . the energy liberated by accretion has to escape . since the accretion flow is rotating , some of it could exhaust through a low - density funnel . but much of it presumably percolates through the accretion flow and envelope . thus , the accreting black hole provides an energy source for the envelope , which is therefore a kind of star - like object which we have dubbed a quasistar " ( begelman et al . 2006 ) . since there is a limit to the outward energy flux that the accreting gas can carry , the accretion rate is regulated at a fraction @xmath42 of the bondi rate , where @xmath43 is the sound speed at the bondi radius ( gruzinov 1998 ; blandford & begelman 1999 ; narayan et al . 2000 ; quataert & gruzinov 2000 ) . the accretion rate , and associated energy flux , can be expressed in terms of the temperature and density deep within the quasistar , but outside the black hole s sphere of influence . the energy flux must also equal the eddington limit for the entire quasistar , @xmath44 ( where we now use @xmath45 to denote the mass of the quasistar ) the interior structure adjusts so that this is satisfied . since the eddington limit is lower inside the quasistar ( where the enclosed mass is lower ) , the quasistar s interior is strongly convective and can be modeled by an @xmath46 polytrope . ( strictly speaking , one should use a loaded polytrope " model [ huntley & saslaw 1975 ] , taking into account the mass of the black hole , but the loaded and unloaded models converge in the region of interest when @xmath47 . ) the interior structure of a quasistar is shown schematically in fig . 1 . accretes gas from a massive , radiation pressure - supported envelope at a rate set by conditions outside the bondi radius . the luminosity liberated by the accretion process is transported convectively in the inner regions of the envelope , with a transition to a radiative zone where convection becomes inefficient . we illustrate rotational flattening and ongoing disk accretion at a fraction of a solar mass per year . ] according to the polytropic relations , the density and temperature outside the bondi radius are uniquely related to the mass and radius ( @xmath48 ) of the quasistar . expressing the accretion rate in terms of these quantities and setting the luminosity equal to @xmath44 , we can express @xmath48 in terms of @xmath45 and @xmath49 : @xmath50 where the masses are expressed in solar units and @xmath51 parametrizes the efficiency of angular momentum and energy transport inside the bondi radius ( begelman et al . the photospheric temperature is @xmath52 since the black hole growth rate is proportional to @xmath45 , while the quasistar mass increases linearly with time , @xmath53 at late times . this implies that the quasistar s radius grows with time , and its photosphere becomes cooler . ( the result holds even if the quasistar mass is fixed . ) the appearance of a quasistar differs dramatically from that of the pre - black hole envelope . shortly after the black hole forms the envelope expands to @xmath54 au ( from @xmath55 au ) , and the interior temperature drops to @xmath56 k , quenching all nuclear reactions . a quasistar resembles a red supergiant , except that it is radiation pressure supported and its energy source is accretion . in some respects it is reminiscent of a very massive thorne - ytkow ( 1977 ) object , but with crucial differences . besides being dominated by radiation pressure and powered exclusively by accretion , quasistars have a distributed rather than shell - like energy source , which can not be regulated by the slow settling that characterizes accretion onto a neutron star . as a result , quasistars come to grief if their photospheric temperatures get too low . , the darker shaded region is for @xmath57 , while the dashed line is for an analytic toy " opacity model . superimposed on the figure are evolutionary tracks for an accretion rate onto the envelope of @xmath58= 1 @xmath59 and @xmath60= 0.1@xmath61 . the evolution pushes the quasistar into the forbidden region of parameter space , where it evaporates . ] begelman et al . ( 2007 ) computed quasistar models using pop iii opacities from mayer & duschl ( 2005 ) . as in red giants with standard abundances , there is a minimum photospheric temperature associated with a sharp drop in opacity . because of the absence of metals , the opacity crisis " occurs at a somewhat higher temperature around 4000 k than for metal - enriched compositions . like the hayashi track ( hayashi 1961 ; hayashi & hoshi 1961 ) for red giants and convective protostars , the minimum temperature of quasistars arises from the impossibility of matching the convective interior to the radiative zone and photosphere . the details are somewhat different , however , because radiation pressure dominates in quasistars , whereas the convective zones of ordinary red giants are gas pressure - dominated and resemble @xmath62 polytropes . as a quasistar crosses into the forbidden zone , the flux escaping from the convective interior exceeds the eddington limit and prevents the quasistar from maintaining hydrostatic equilibrium . 2 shows the forbidden zone in the @xmath63 plane , along with representative evolutionary tracks . unlike ordinary red giants reaching the hayashi track , quasistar photospheres can not hover stably at close to the minimum temperature . fixing the photospheric temperature , we find that the eddington ratio scales as @xmath64 thus , growth of the black hole exacerbates the dynamical imbalance , as does partial evaporation of the envelope , rapid accretion onto the envelope also would not be able to stabilize the quasistar for very long . ] the opacity crisis is an unstable situation . we predict that quasistars entering the forbidden zone must evaporate rather quickly . there is a complication due to a bump in the opacity from bound - free transitions , which creates a narrow region in the radiative zone where the luminosity is super - eddington . a similar problem arises in models of luminous blue variables ( e.g. , owocki et al . 2004 ) , and remains unresolved . however , it does not seem likely that this feature alone will lead to catastrophic mass loss in the case of quasistars . the rapid infall of gas in galactic nuclei or pregalactic haloes provides a means for forming seed black holes and rapidly growing them into the intermediate mass regime . for pop iii quasistars with inflow rates @xmath65 yr@xmath5 , and simple assumptions about parameters like @xmath66 , the black holes could reach masses @xmath67 before the quasistar evaporates . the quasistar masses could be as large as @xmath68 . metal - rich quasistars could reach somewhat lower temperatures , but it is not clear whether this implies larger black hole masses , because the run of opacity in the radiative zone would be more complex and dust formation might lead to enhanced mass loss . are quasistars detectable ? if so , they would be seen at their most massive , shortly before they evaporate . spectrally , they would resemble 20004000 k blackbodies , depending on metallicity , and in the pop iii limit their spectra would be featureless . since they radiate at the eddington limit corresponding to the opacity at the convective radiative transition which is close to that of electron scattering their luminosities could reach @xmath69 erg s@xmath5 . however , their short lifetimes ( @xmath70 yr ) would make them fairly rare . quasistars can exist only when the envelope mass greatly exceeds the mass of the embedded black hole . therefore , they are unlikely to form in the nuclei of galaxies that already possess a supermassive black hole in this case , a period of rapid infall ( e.g. , following a merger ) would presumably trigger a quasar outburst instead . however , low - redshift quasistars might conceivably form in galactic nuclei in which the black hole had been ejected due to three - body interactions , or in which a black hole had never formed . the earliest plausible sites of quasistar formation would have been pregalactic haloes with virial temperatures exceeding @xmath71 k. these would have been most common at redshifts @xmath72 . the spectra of pop iii quasistars at these redshifts would peak at about 10@xmath73 m , but they might be marginally detectable by _ james webb space telescope _ on the wien tail at @xmath74 m . if direct collapse in @xmath71 k haloes is a principal route for forming supermassive black hole seeds , there could be as many as 110 per _ jwst _ field , but identifying them would be an extreme challenge . the work described here was done in collaboration with marta volonteri , elena rossi , martin rees , and phil armitage , and was supported in part by nsf grant ast-0307502 , nasa s beyond einstein foundation science program grant nng05gi92 g , and a university of colorado faculty fellowship . the author thanks the director of the institute of astronomy and the master and fellows of trinity college , cambridge , for their hospitality during a sabbatical visit .	rapid infall of gas in the nuclei of galaxies could lead to the formation of black holes by direct collapse , without first forming stars . black holes formed in this way would have initial masses of a few @xmath0 , but would be embedded in massive envelopes that would allow them to grow at a highly super - eddington rate . thus , seed black holes as large as @xmath1 could form very rapidly . i will sketch the basic physics of the direct collapse process and the properties of the accreting envelopes . address = jila , 440 ucb , university of colorado at boulder , boulder , co 80309 - 0440 usa
the physical relevance of scalar fields in today s gravitational physics and cosmology primarily stems from i ) their key role in current models of early cosmological inflation , with predictions that to an astonishing degree have been confirmed by recent cosmological measurements @xcite , and ii ) the viability of scalar field models as candidate models for dark energy @xcite@xcite@xcite . the fact that scalar fields are inevitable artifacts of string theory @xcite provides additional rationale for studying scalar fields . the continuing focus on extra - dimensional models in fundamental physics provides motivation for studying scalar fields in dimensions higher than 4 . there is an extensive literature on known einstein - scalar field solutions . in a recent paper , wehus and ravndal @xcite provide a good historical overview of the various einstein - scalar field solutions relevant for the present paper . solution generation is a powerful method of discovering new solutions and uncovering relationships between different solutions . there are numerous examples of solutions to the einstein - scalar field equations , which were uncovered using solution generation techniques , that would be extremely hard to derive by other means , including solving the field equations by brute force . the way these solutions typiclly have been derived is by first finding a new solution generation technique that transforms a class of vacuum solutions into a class of einstein - scalar field solutions . subsequently , this technique is applied to a known vacuum solution of the einstein equations to generate a new solution to the einstein - scalar field equations . techniques for generating both static as well as time dependent solutions are known by know . buchdahl @xcite derived static solutions to the minimally coupled einstein - scalar field equations in 4 dimensions by applying a particular transformation to vacuum solutions of the einstein equations . he applied the solution generation technique to the schwarzschild solution and obtained the general static , spherically symmetric solution to the minimally coupled einstein - scalar field equations for a massless scalar field . this particular solution had been derived earlier by solving the field equations directly , first by bergmann and leipnik @xcite and later by janis , newman and winicour @xcite . the buchdahl transformation was later rediscovered by janis , robinson and winicour @xcite . the static , spherically symmetric scalar field solutions in 4 dimensions were generalized to arbitrary dimension by xanthopoulos and zannias @xcite by solving the einstein - scalar field equations in arbitrary dimension . the action for a scalar field conformally coupled to einstein gravity in 4 dimensions was formulated by callan , coleman and jackiw @xcite . bocharova , bronnikov and melnikov found a static , spherically symmetric black hole solution to the conformally coupled einstein - scalar field equations by directly solving the field equations . this solution , however , was not known in the west until much later . independently , bekenstein @xcite found a solution generation technique that from an arbitrary solution to the massless , minimally coupled scalar field equations in 4 dimensions could generate two solutions to the massless , conformally coupled scalar field equations . he first applied the buchdahl transformation to the scwharzschild solution and then his solution generation technique to generate conformally coupled , spherically symmetric solutions . the black hole solution of bocharova et . al . is among the generated solutions . this solution is now known as the bbmb scalar black hole solution . the bekenstein solutions , including the bbmb black hole , were later rediscovered by fryland @xcite . maeda generalized the bekenstein solution technique to arbitrary dimension @xcite . the bekenstein technique is very general , and applies to any solution to the minimally coupled scalar field equations . husain , martinez and nnez found a time dependent , spherically symmetric solution to the minimally coupled scalar field equations with zero scalar potential @xcite . this solution was interpreted as a model of scalar field collapse . soon after , fonarev found a technique for generating time dependent solutions to the einstein - scalar field equations in 4 dimensions from static vacuum solutions to the einstein equations @xcite that extended the husain et . al solution to non - vanishing scalar potential . by combining his technique with the bekenstein transformation , he identified a time dependent , conformally coupled solution that describes the formation of a scalar black hole . feinstein , kunze and vzquez - mozo later generalized the fonarev solutions to five dimensions @xcite . in this paper , we will focus our attention on solutions to the minimally coupled einstein - scalar field equations in arbitrary dimension . we will identify a class of solutions that includes both the buchdahl as well as the fonarev solutions . this class of solutions can then be combined with the bekenstein transformation in order to generate conformally coupled scalar field solutions . in section [ xref - section-5159214 ] , we review the minimally coupled scalar field action and the einstein - scalar field equations . section [ xref - section-421172432 ] reviews the general form of the vacuum solutions we will use as seed for the transformation . in section [ xref - section-42223153 ] , we provide a brief review of the buchdahl transformation in 4 dimensions . section [ xref - section-51592326 ] introduces the classification scheme that later will be used to label the different solutions . section [ xref - section-51592441 ] contains our main results , as it proves the new solution transformation . the section ends by providing the new solution transformation in a unified form that encapsulates both the buchdahl and fonarev transformations . finally , in section [ xref - section-5159268 ] , we study an example solution of the new kind , derived from an @xmath0 vacuum solution . there are two appendices : appendix [ xref - appendixsection-98225849 ] contains expressions that are useful for conformal transformations . appendix [ xref - appendixsection-510235039 ] provides , for refrence purposes , the generalized bekenstein theorem in arbitrary dimension and for arbitrary scalar potential . this paper deals with solution generation techniques for minimally coupled scalar fields . the solutions sought are solutions to the minimally coupled coupled einstein - scalar field equations . let us briefly review the einstein - scalar field action and the field equations derived from it when the action is extremalized . in @xmath3 dimensions , the action for a scalar field minimally coupled to gravity is @xmath4 = \int d^{n}x\sqrt{-g}\left ( \frac{1}{2\kappa } \left ( r [ g ] -2\lambda \right ) -\frac{1}{2}g^{\alpha \beta } \nabla _ { \alpha } \phi \nabla _ { \beta } \phi -v [ \phi ] \right ) , \ ] ] where @xmath5 is a cosmological constant , @xmath6 , @xmath7 being the @xmath3dimensional gravitational constant and @xmath8 $ ] is an arbitrary scalar potential of the scalar field @xmath9 . hereafter , we will include @xmath5 in the potential term , defining an effective potential @xmath10 \equiv v [ \phi ] + \kappa ^{-1}\lambda $ ] . we get the minimally coupled einstein - scalar field equations by extremalizing the above action : @xmath11 -\frac{1}{2}g_{\mu \nu } r [ g ] = \kappa \left ( \nabla _ { \mu } \phi \nabla _ { \nu } \phi -\frac{1}{2}g_{\mu \nu } g^{\alpha \beta } \nabla _ { \alpha } \phi \nabla _ { \beta } \phi -g_{\mu \nu } w [ \phi ] \right ) % \label{xref - equation-9855237 } \\\box \phi -\frac { d w [ \phi ] } { d \phi } = 0.% \label{xref - equation-9862429}\end{gathered}\ ] ] @xmath12 is the covariant derivative associated with the metric @xmath13 , and @xmath14 is the @xmath3 dimensional laplace operator of this metric . @xmath15 $ ] and @xmath16 $ ] are the ricci tensor and ricci scalar for an arbitrary metric @xmath13 . by contracting equation [ xref - equation-9855237 ] , we can rewrite this equation as @xmath17 = \kappa ( \nabla _ { \mu } \phi \nabla _ { \nu } \phi + \frac{2}{n-2}g_{\mu \nu } w [ \phi ] ) % \label{xref - equation-9862442}\ ] ] all solution transformations derived or referenced in this paper use vacuum solutions as seed . by vacuum solution , we mean a solution to the vacuum einstein equations with an arbitrary cosmological constant @xmath5 . we will require the seed solutions to possess a hypersurface - orthogonal killing vector field @xcite , in which case it is possible to choose coordinates for which the line element associated with the metric @xmath13 for the this geometry takes the following canonical form : @xmath18 = \epsilon e^{2u ( x ) } { \mathrm{dy}}^{2}+e^{-2u ( x ) } h_{\mathrm{ij}}{\mathrm{dx}}^{i}{\mathrm{dx}}^{j}.% \label{xref - equation-91972716}\ ] ] the coordinate @xmath19 is a parameter of the integral curves of the killing vector field , @xmath20 or @xmath21 , depending on whether the killing vectors are time - like or space - like , @xmath22 is a scalar function that is independent of the @xmath19-coordinate and @xmath23 is the metric of a foliation of @xmath24 dimensional hypersurfaces orthogonal to the y - direction with coordinates @xmath25 . in the following , we will use @xmath26 $ ] to denote a metric written on the form of eq . [ xref - equation-91972716 ] : @xmath27 = \epsilon e^{2u ( x ) } , g_{\mathrm{ij } } [ u , h ] = e^{-2u ( x ) } h_{\mathrm{ij } } , g_{0i } [ u , h ] = 0% \label{xref - equation-92192055}\ ] ] throughout the paper , we will refer to this type of geometry as a geometry with an _ orthogonal symmetry _ or being _ orthogonal - symmetric_. likewise , we will refer to a metric of the form of eq . [ xref - equation-92192055 ] as a metric with an _ orthogonal symmetry _ or an _ orthogonal - symmetric _ metric . several well - known static vacuum solutions to the einstein equations are orthogonal - symmetric , including the schwarzschild solution and the weyl solutions . the kerr solution is an example of a space - time geometry that is not orthogonal - symmetric . although the kerr solution is stationary , which means that it has a time - like killing vector field , its @xmath28 hypersufaces are not orthogonal to the time - like killing vector , and its metric can not be put on the canonical form of eq . [ xref - equation-91972716 ] . an important class of orthogonal - symmetric geometries is the class of geometries with a time - like killing vector orthogonal to a set of space - like hypersurfaces . these are static geometries , which posess metrics on the canonical form : @xmath29 both the buchdahl and fonarev transformations were derived using static seed solutions on this form @xcite . because it will become useful in later derivations , let us evaluate the ricci tensor of an arbitrary orthogonal - symmetric metric written on the form of equation [ xref - equation-91972716 ] in terms of the scalar function @xmath22 , the @xmath24 dimensional metric @xmath30 and its corresponding ricci tensor @xmath31 $ ] . a lenghty calculation reveals the following expressions for the ricci tensor : @xmath32 = -\epsilon e^{4u } ( \nabla ^{2}u-\left ( n-4\right ) h^{\mathrm{kj}}\partial _ { k}u \partial _ { j}u ) % \label{xref - equation-92082543 } \\r_{0i } [ g ] = 0% \label{xref - equation-4247154 } \\r_{\mathrm{kj } } [ g ] = h_{\mathrm{kj}}-\left ( 6-n\right ) \nabla _ { j}u \nabla _ { k}u+h_{\mathrm{kj } } ( \nabla ^{2}u-\left ( n-4\right ) h^{\mathrm{il}}\partial _ { i}u\ \ \partial _ { l}u ) + \left ( n-4\right ) \nabla _ { k}\nabla _ { j}u,% \label{xref - equation-9276244}\end{gathered}\ ] ] where @xmath33 . similarly , we may express the @xmath3 dimensional laplacian operator @xmath34 in terms of the @xmath24 dimensional laplacian @xmath35 @xmath36 before we treat the general case , let us briefly review the buchdahl transformation @xcite . starting with a static vacuum solution to the einstein equations with vanishing cosmological constant in 4 dimensions , the metric @xmath37 $ ] can be written on the form of eq . [ xref - equation-91972716 ] : @xmath38 = -e^{2v ( x ) } { \mathrm{dt}}^{2}+e^{-2v ( x ) } h_{\mathrm{ij}}{\mathrm{dx}}^{i}{\mathrm{dx}}^{j},\ ] ] where @xmath39 is a scalar function of the spatial coordinates @xmath40 . we will seek static solutions to the massless einstein - scalar field equations , eqs . [ xref - equation-9862429 ] and [ xref - equation-9862442 ] with metrics on this form . since @xmath41 is the metric of a vacuum solution to the einstein equations , it follows from eq . [ xref - equation-92082543 ] that @xmath42 must satisfy the 3 dimensional laplace equation , @xmath43 . furthermore , by applying eq . [ xref - equation-9276244 ] , we find that @xmath44 now , let us write the einstein - scalar field equations eqs . [ xref - equation-9862429 ] and [ xref - equation-9862442 ] for a general orthogonal - symmetric solution . the scalar field equation , eq . [ xref - equation-9862429 ] , takes the following form , using eq . [ xref - equation-9208260 ] and that we are restricting our attention to a scalar field with zero potential : @xmath45 the einstein equations take the following form , using eqs . [ xref - equation-92082543 ] and [ xref - equation-9276244 ] and assuming a static scalar field : @xmath46 asserting that @xmath47 and @xmath48 both are functionals of @xmath42 , i.e. @xmath49 $ ] and @xmath50 $ ] , we can rewrite eqs . [ xref - equation-42115556 ] and [ xref - equation-421155516 ] : @xmath51 \nabla ^{\alpha } v\nabla _ { \alpha } v+k^{\prime } [ v ] \nabla ^{2}v=0 \\h_{\mathrm{kj}}-\left ( 2{k^{\prime } [ v ] } ^{2}+\kappa { f^{\prime } [ v ] } ^{2}\right ) \nabla _ { j}v\nabla _ { k}v=0\end{gathered}\ ] ] using that that @xmath42 satisfies the laplace equation @xmath43 as well as eq . [ xref - equation-42116542 ] , we get the following constraints for the functionals @xmath52 and @xmath53 : @xmath54 = 0 \\\kappa { f^{\prime } [ v ] } ^{2}+2{k^{\prime } [ v ] } ^{2}=2\end{gathered}\ ] ] this gives @xmath55 where @xmath56 , @xmath57 and @xmath58 are arbitrary constants and @xmath59 is a constant that satisfies @xmath60 . @xmath57 just gives a scale factor that can be absorbed by rescaling the coordinates , so we may set @xmath61 . assuming that @xmath13 is asymptotically flat , @xmath58 is the asymptotic value of the field @xmath62 , so we will set @xmath63 in order to let the scalar field @xmath62 go to zero as @xmath64 at infinity . for an arbitrary static vacuum solution @xmath65 $ ] to the 4 dimensional einstein equations , ( @xmath66 $ ] , @xmath62 ) is a solution to the minimally coupled einstein - scalar field equations with zero scalar potential : @xmath67 where @xmath68 is an arbitrary dimensionless constant in the interval @xmath69 $ ] , and @xmath70 [ xref - equation-421163941 ] and [ xref - equation-421163952 ] define the buchdahl soluions , and the buchdahl transformation is defined by the relationships between the scalar field @xmath62 and metric functions @xmath48 and @xmath42 given by the functionals @xmath52 and @xmath53 . we see that the buchdahl transformation is straight forward to prove in 4 dimension . in other dimensions , this is , as we will see later , no longer the case . in this section we will introduce the general classification scheme that will be used throughout the paper to label different solutions to the einstein equations . the classification scheme is motivated by the new solution transformation technique that will be derived in the next section , and it labels scalar field solutions by characteristics of the seed solution as well as properties of the transformation applied to the seed solution . the new solution transformation is a rather general result that encompasses two known solution transformation techniques in 4 dimensions , the buchdahl @xcite and fonarev @xcite transformations . the solution transformations referenced and derived in this paper all share a common characteristic ; they all derive minimally coupled einstein - scalar field solutions from orthogonal - symmetric vacuum solutions to the einstein equations , which according to the definition of orthogonal - symmetric fields introduced in section [ xref - section-421172432 ] implies that the geometry of the seed solution must posesses a hypersurface - orthogonal killing vector field and , subsequently , the solution is invariant with respect to translations along the integral curves of the killing vector field . the buchdahl and fonarev solution transformations were derived for the special case of static seed solutions with timelike , hypersurface - orthogonal killing vector fields , but the solution transformations extend easily to the more general case of arbitrary hypersurface - orthogonal spacetime symmetries . as seen in the previous section , the buchdahl transformation @xcite transforms an arbitrary 4-dimensional static vacuum solution @xmath71 $ ] into a one - parameter family of static , minimally coupled scalar field solutions with metrics @xmath72 , h ] $ ] . the metric functions @xmath73 = \alpha v$ ] is parametrized by a a dimensionless parameter @xmath68 in the range @xmath69 $ ] . the fonarev transformation @xcite extends the buchdahl transformation . it acts on the same class of seed vacuum solutions as the buchdahl transformation , which are solution metrics fitting the metric template of eq . [ xref - equation-91972716 ] , but the resulting scalar field solutions are time - dependent with metrics that differ from eq . [ xref - equation-91972716 ] by a time - dependent conformal factor : @xmath74 we will refer to metrics that fit this metric template as _ conformally inflated metrics _ , or to be more specific , _ conformally time - inflated metrics , _ referring to the time - dependent conformal factor that transforms a static space - time geometry into an inflating / deflating one . the geometry of eq . [ xref - equation-9266298 ] is clearly conformally static . let us generalize this geometry to a geometry that is conformally ortogonal - symmetric : @xmath75 the metric templates of eqs . [ xref - equation-91972716 ] and [ xref - equation-42218146 ] motivate the introduction of a solution classification based on how the generated solutions fit the template of eq . [ xref - equation-42218146 ] . orthogonal - symmetric solutions with @xmath76 represent one class of generated solutions , hereafter given the label @xmath77 , whereas another class are the @xmath19-dependent solutions with@xmath78 , hereafter given the label @xmath79 . class @xmath80 solutions are hereafter referred to as _ conformally y - inflated _ solutions , referring to the y - dependent conformal factor @xmath81 that represents an inflation / deflation of a y - invariant geometry along the y - direction . the buchdahl and fonarev transformations share a vital restriction : any seed solution for these transformations must be a vacuum solution to the einstein equations with vanishing cosmological constant ( @xmath82 ) . if we are able to lift this restriction , i.e. derive scalar field solutions from vacuum solutions with non - vanishing cosmological constant , we get two new classes of solutions : * @xmath83 : orthogonal - symmetric solutions ( with @xmath76 ) , derived from orthogonal - symmetric vacuum solutions with @xmath84 * @xmath85 : @xmath19-dependent , conformally y - inflated solutions derived from orthogonal - symmetric vacuum solutions with @xmath86 @xmath87 0 to be consistent with this labeling , we will label the solution classes @xmath88 and @xmath79 introduced above as @xmath89 and @xmath90 , respectively , indicating that they derive from seed solutions with @xmath91 . let us summarize our scheme for labeling minimally coupled scalar field solutions derived from vacuum solutions to the einstein equations : * @xmath92 : orthogonal - symmetric solutions ( @xmath76 ) derived from orthogonal - symmetric vacuum solutions with @xmath86 @xmath87 0 * @xmath93 : orthogonal - symmetric solutions ( @xmath94 ) derived from orthogonal - symmetric vacuum solutions with @xmath86 = 0 * @xmath95 : conformally y - inflated solutions ( @xmath96 ) derived from orthogonal - symmetric vacuum solutions with @xmath86 @xmath87 0 * @xmath97 : conformally y - inflated solutions ( @xmath98 ) derived form orthogonal - symmetric vacuum solutions with @xmath86 = 0 from equation [ xref - equation-9862442 ] we see that vacuum solutions to the einstein equations must satisfy the equation @xmath17 = \frac{2}{n-2}g_{\mu \nu } \lambda , % \label{xref - equation-92783053}\ ] ] with @xmath5 being an arbitrary cosmological constant . we will take an arbitrary orthogonal - symmetric vacuum solution to the einstein equations , @xmath99 $ ] , as our seed solution . using eq . [ xref - equation-91972716 ] , the metric @xmath41 of this solution can be written on the canonical form @xmath100 we are seeking to find the most general solution to the minimally coupled einstein - scalar field equations , eqs . [ xref - equation-9862429 ] and [ xref - equation-9862442 ] , matching the metric template of a conformally y - inflated metric : @xmath101 where the @xmath24 dimensional metric @xmath102is conformally related to @xmath103 by a y - independent conformal transformation : @xmath104 and posessing a scalar field on the form @xmath105 with scalar field potential @xmath106 $ ] . for the time being , we assume nothing about the potential , so it can be any function of the scalar field . in the above expressions , we have assumed that @xmath107 and @xmath108 are parameters that will be subject to constraints in order to provide solutions to the field equations . before we proceed , let us relate the solution template defined by equations [ xref - equation-926225147]-[xref - equation-92962746 ] to known solutions . for example , we can retrieve the buchdahl solutions @xcite in 4 dimensions , labeled @xmath109 , by setting @xmath110 and @xmath5 to zero . deriving our results in arbitrary dimension , we should therefore immediately be able to generalize the buchdahl transformation from 4 to arbitrary dimension , i.e. derive solutions of class @xmath93 . furthermore , if we set @xmath5 to zero , but allow @xmath110 to vary with @xmath19 , we retrieve the fonarev class of solutions@xcite in 4 dimensions , labeled @xmath111 . like for the buchdahl transformation , we should be able to generalize the fonarev transformation to arbitrary dimension , i.e. derive solutions of class @xmath97 . finally , the metric template of equation [ xref - equation-926225147 ] covers several classes of solutions that are not covered by previously published results : * @xmath92 : orthogonal - symmetric solutions derived from vacuum solutions with non - zero cosmological constant * @xmath95 : y - dependent , conformally inflated solutions derived from vacuum solutions with non - zero cosmological constant let @xmath112 denote the metric of equation [ xref - equation-926225147 ] with @xmath113 being an orthogonal - symmetric metric ; @xmath114 $ ] . in the following , we will use overdots to represent derivatives wrt . the coordinate y. applying the general conformal transformation formulas of appendix [ xref - appendixsection-98225849 ] , we get the following expressions that relate the ricci tensors of the metrics g and @xmath113 : @xmath115 @xmath116 transforms as follows under a y - dependent conformal transformation : @xmath117 we will now proceed by first evaluating the ricci tensor @xmath118using equations [ xref - equation-9953847]-[xref - equation-9953857 ] . @xmath119can be derived from equations [ xref - equation-92082543 ] - [ xref - equation-9276244 ] . when evaluating @xmath120 , it is convenient to isolate terms containing @xmath121 and use the fact that it is a vacuum solution to the einstein equations in order to reduce the expressions as much as possible . having evaluated the ricci tensor , we apply it to the einstein equations , eq . [ xref - equation-9862442 ] . in the course of development of the equations , it is useful to expand the @xmath3-dimensional ricci tensor of an arbitrary orthogonal - symmetric metric @xmath122 $ ] , @xmath123 ] $ ] , into expressions involving the metric function @xmath48 , the metric @xmath24 dimensional metric @xmath124 and its ricci tensor @xmath125 $ ] . equations [ xref - equation-92082543 ] - [ xref - equation-9276244 ] can be used to do that . the following expression for the laplacian @xmath126 of an @xmath19-dependent function @xmath110 is also useful : @xmath127 after a lot of algebra , we find that the @xmath128 component of the einstein equations , eq . [ xref - equation-9862442 ] , reduces to @xmath129 - \alpha e^{-2\left ( 1 - 2\alpha -\delta \right ) v}\lambda \right ) % \label{xref - equation-92963758}\end{gathered}\ ] ] and the @xmath130 component of the einstein equations reduces to @xmath131 -\frac{4 - 2\alpha -2 \delta } { n-2}\lambda \ \ e^{-2\left ( 1-\delta \right ) v } + \epsilon \left ( \ddot{\mu } + \left ( n-2\right ) { \dot{\mu } } ^{2}\right ) e^{-4\alpha v}\right ) h_{\mathrm{kj}}% \label{xref - equation-10495730}\end{gathered}\ ] ] we see that in order for equation [ xref - equation-10495730 ] to be satisfied with @xmath132 , we must have @xmath133 the ( 0,0 ) component of the einstein equations then becomes @xmath134 - \alpha \lambda e^{-2\left ( 1 - 2\alpha -\delta \right ) v}\right ) % \label{xref - equation-424183122}\ ] ] furthermore , since the left - hand side of equation [ xref - equation-92963758 ] depends on @xmath19 only , it is evident that the only way to satisfy this equation when @xmath135 and @xmath136 is to demand @xmath137 . since eq . [ xref - equation-92964657 ] also must be satisfied , we find that when @xmath135 and @xmath138 , @xmath68 must take the unique value @xmath139 . from eq . [ xref - equation-4247154 ] we find that @xmath140 for an orthogonal - symmetric metric . therefore , the @xmath141 component of eq . [ xref - equation-9862442 ] is identically satisfied if @xmath142 . if @xmath143 , it gives us the parametric constraint @xmath144 finally , the @xmath130 component of equation [ xref - equation-9862442 ] becomes , having used equation [ xref - equation-92964657 ] to reduce it : @xmath145 -\frac{4 - 2\alpha -2 \delta } { n-2}\lambda \ \ e^{-2\left ( 1-\delta \right ) v } + \epsilon \left ( \ddot{\mu } + \left ( n-2\right ) { \dot{\mu } } ^{2}\right ) e^{-4\alpha v}\right ) h_{\mathrm{kj}}\end{gathered}\ ] ] each side of this equation must vanish identically , so we get another parametric constaint @xmath146 and the reduced equation @xmath147 + \epsilon \frac{4 - 2\alpha -2 \delta } { n-2}\lambda \ \ e^{-2\left ( 1-\delta -2\alpha \right ) v}% \label{xref - equation-9297151}\ ] ] solving equation [ xref - equation-1097454 ] for @xmath148 , using equation [ xref - equation-92964657 ] , gives @xmath148 in terms of @xmath68 : @xmath149 this is the same relation we found for the buchdahl transformation in section [ xref - section-42223153 ] in 4 dimensions . now , if we turn our attention to the equation of motion for the scalar field , equation [ xref - equation-9862429 ] , we can use equation [ xref - equation-92763148 ] to compute @xmath150 . the scalar field equation , eq . [ xref - equation-9862429 ] , then takes the form @xmath151 + \epsilon \ \ \frac{2}{n-2}\ \ \frac{\sigma } { \rho } \lambda e^{-2\left ( 1 - 2\alpha -\delta \right ) v}% \label{xref - equation-929720}\ ] ] we immediately see that the the left - hand sides of equations [ xref - equation-9297151 ] and [ xref - equation-929720 ] are identical , and that , by equating the right - hand sides of these equations , we get an equation constraining the scalar field potential : @xmath152 = -\frac{2\kappa \rho } { n-2}w [ \phi ] + \frac{2 \rho } { n-2}\left ( 2-\alpha - \delta -\frac{\sigma } { \rho } \right ) \lambda \ \ e^{2\left ( \alpha + \delta -1\right ) \frac{1}{\sigma } \phi } e^{-2\left ( \left ( \alpha + \delta -1\right ) \frac{\rho } { \sigma } + 1\right ) \mu } % \label{xref - equation-10617136}\ ] ] one way in which to satisfy equation [ xref - equation-10617136 ] is to have the second term on the right - hand side of this equation vanish , which may happen when the following equation is satisfied : @xmath152 = -\frac{2\kappa \rho } { n-2}w [ \phi ] % \label{xref - equation-1037126}\ ] ] and either @xmath153 or @xmath82 ( assuming @xmath154 ) . we will refer to this as case 1 . equation [ xref - equation-1037126 ] can easily be integrated , and gives the exponential scalar field potential @xmath155 = w_{0 } e^{-2k \phi } , % \label{xref - equation-106211723}\ ] ] where @xmath156 and @xmath157 is an arbitrary constant . notice that , although equation [ xref - equation-1037126 ] admits an additional constant term in the potential , equations [ xref - equation-9297151 ] and [ xref - equation-929720 ] will not be identical unless this constant term vanishes . another way in which equation [ xref - equation-10617136 ] can be satisfied is that the @xmath110-dependence in the last term vanishes , either by simply having @xmath76 ( case 2 ) , or by setting @xmath158 ( case 3 ) . evidently , it is not sufficient to require that equations [ xref - equation-9297151 ] and [ xref - equation-929720 ] are identical ; the equations must also admit solutions . this adds more parametric constraints , because it means that , provided @xmath159 we must choose parameters that makes the @xmath42-dependence of both eqs . [ xref - equation-424183122 ] and [ xref - equation-9297151 ] vanish . we will now show that equation [ xref - equation-106211723 ] is the unique scalar field potential that satisfies equation [ xref - equation-10617136 ] and is not in conflict with neither the constraints nor the @xmath128 field equation , eq . [ xref - equation-424183122 ] . this will prune our solution space considerably , as we will have a unique expression to use for the scalar field potential . the simplest case is case 2 , i.e. @xmath76 . in this case , we are able to derive the potential directly from equation [ xref - equation-9297151 ] . inserting this expression in equation [ xref - equation-10617136 ] gives further parametric constraints . when inserting the solutions into the ( 0,0 ) field equation , eq . [ xref - equation-424183122 ] , we find that the only solution admitted in this case is the potential of equation [ xref - equation-106211723 ] with @xmath91 . however , if both @xmath91 and @xmath160 , eq . [ xref - equation-9297151 ] demands @xmath161 . finally , considering case 3 , we find the same thing . the potential must have the exponential form of equation [ xref - equation-106211723 ] , and any case 3 solution with @xmath136 is inconsistent with the @xmath128 field equation . let us summarize what we have learnt so far : given an orthogonally symmetric vacuum solution , @xmath162g[v,@xmath163 , to the n dimensional einstein equations wth an arbitrary cosmological constant , equation [ xref - equation-92783053 ] , a scalar field on the form [ xref - theorem-103203143 ] @xmath164with a scalar field potential @xmath155 = w_{0 } e^{-2k \phi } \]]and a conformally y - inflated space - time with a metric on the form @xmath165is a solution to the minimally coupled einstein - scalar field equations , equations [ xref - equation-9862429 ] and [ xref - equation-9862442 ] , if and only if the following equations are satisfied : @xmath166where @xmath86 is an arbitrary cosmological constant and @xmath68 , @xmath167 , @xmath148 , @xmath168 , k , and @xmath157 are parameters satisfying the following set of constraints : @xmath169if @xmath170 and @xmath91 , the parameter @xmath167 is given by @xmath171if @xmath172 , the parameter k of the scalar field potential is given in terms of @xmath167 : @xmath173if @xmath174@xmath870 , the following constraint must be satisfied : @xmath175if @xmath5@xmath870 , we must have @xmath176which has the unique solution @xmath139 . equations [ xref - equation-103193029]-[xref - equation-103193044 ] do indeed cover a wide range of scalar field solutions , and we will use the rest of this section to unravel the main solution classes . as a result , we will be able to accurately label the einstein - scalar field solutions that can be generated by the solution transformation of theorem [ xref - theorem-103203143 ] in terms of the labeling scheme introduced in the previous section . before we proceed , let us review the different cases to be considered : * @xmath76 ( solution class @xmath177 ) . * * @xmath86 @xmath870 ( solution class @xmath178 ) * * @xmath86=0 ( solution class @xmath93 ) * @xmath179 ( solution class @xmath180 ) * * @xmath86 = 0 ( solution class @xmath97 ) * * @xmath86@xmath870 ( solution class @xmath181 ) the various constraints set up by theorem [ xref - theorem-103203143 ] relate to the different solution classes as follows : all solutions must satisfy equations [ xref - equation-10322143 ] and [ xref - equation-10321522 ] . all class @xmath79 solutions with non - vanishing scalar potential must in addition satisfy equation [ xref - equation-103193044 ] . @xmath92and @xmath95solutions must satisfy equations [ xref - equation-103214959 ] and [ xref - equation-10982211 ] , while any solution with a non - vanishing scalar field potential must satisfy equation [ xref - equation-1098246 ] . let us consider for a moment the case of a seed vacuum solution with constant metric function @xmath42 . without loss of generality , we can set @xmath42 to zero . from equations [ xref - equation-92082543 ] and [ xref - equation-92783053 ] we immediately see that the @xmath128 component of the einstein equations , eq . [ xref - equation-9862442 ] , requires @xmath5 to be zero in this case , so this case falls in the @xmath93and @xmath97solution classes . in the following two sections , we will state the individual sub classes of @xmath88 and @xmath80 solutions in terms of separate corollaries of theorem [ xref - theorem-103203143 ] . the @xmath177classes of solutions emerge by setting @xmath94 in theorem [ xref - theorem-103203143 ] . in that case , equations [ xref - equation-103193029 ] and [ xref - equation-103214451 ] take the form @xmath183 considering the @xmath92class of solutions first , i.e. assuming @xmath136 , we immediately find the constraint @xmath184 . in this case , equation [ xref - equation-103214959 ] implies that @xmath56 must be @xmath21 . setting @xmath185 into equation [ xref - equation-10321522 ] implies @xmath186 , in conflict with the constraint @xmath187 . furthermore , we have to set @xmath91 in eqs . [ xref - equation-10322628 ] and [ xref - equation-10322635 ] , we must also have @xmath170 . thus , we have proved the following : @xmath188 there are no scalar field solutions of class @xmath189 that can be derived using the transformation of theorem [ xref - theorem-103203143 ] , i.e. there are no orthogonal - symmetric scalar field solution that can be derived from a vacuum solution with cosmological constant @xmath86@xmath87 0 using the solution transformation of theorem [ xref - theorem-103203143 ] . furthermore , @xmath157 must be zero for the entire @xmath177 class of solutions , i .. e there are no solutions of class @xmath177 with a potential term.[xref - corollary-1018173233 ] now , let us turn our attention to the @xmath93solution class , i.e. orthogonal - symmetric scalar field solutions with vanishing cosmological constant . these are the generalized buchdahl solutions in dimension @xmath3 . it is evident from eqs . [ xref - equation-10322628 ] and [ xref - equation-10322635 ] that @xmath157 must also be zero , i.e we are considering the case of a massless scalar field . the remaining constraints in this case are equations [ xref - equation-10322143 ] and [ xref - equation-10321522 ] , with the solution @xmath190 where @xmath191 and @xmath192 are constants given by @xmath193,@xmath194 and @xmath195 . we can immediately see that eq . [ xref - equation-103221949 ] reduces to buchdahl s result of eq . [ xref - equation-421163941 ] for @xmath196 . applying equation [ xref - equation-103221949 ] to theorem [ xref - theorem-103203143 ] allows us to derive the characteristics of the @xmath197solution class : @xmath198 if @xmath162g[v,@xmath163 is an orthogonally symmetric vacuum solution to the einstein equations with vanishing cosmological constant , we have a one - parameter family of solutions to the minimally coupled einstein - scalar field equations for a massless scalar field given by an orthogonal - symmetric space - time metric [ xref - corollary-104611 ] @xmath199with a massless scalar field @xmath200where @xmath68 is a free , dimensionless parameter in the range [ -1,1 ] , @xmath201@xmath202@xmath203 and @xmath148@xmath202@xmath204 . corollary [ xref - corollary-104611 ] was first proved in 4 dimensions for stationary metrics by buchdahl @xcite and later by janis , robinson and winicour @xcite . we refer to this version of the general transformation of theorem [ xref - theorem-103203143 ] as the _ generalized buchdahl transformation_. in 4 dimensions , the buchdahl transformation is easy to prove . however , to our knowledge , this is the first time it has been generalized to arbitrary dimension . notice that for @xmath205 , the generalized buchdahl transformation of corollary [ xref - corollary-104611 ] must do an extra @xmath24 dimensional conformal transformation @xmath206of the @xmath24 dimensional hypersurface in order to provide a valid solution . this implies that the @xmath93scalar field solutions in other dimensions than 4 become , in general , significantly more compelex than the 4 dimensional solutions do . next , we will turn our attention to the @xmath19-dependent solutions , i.e. the @xmath180 solution class , and we will try to generate @xmath19-dependent scalar field solutions from the constraints set up by theorem [ xref - theorem-103203143 ] . as noted above , all class @xmath80 solutions must satisfy the constraints set up by equations [ xref - equation-10322143 ] and [ xref - equation-10321522 ] . @xmath180 solutions with non - vanishing scalar potential , i.e. @xmath172 , must in addition satisfy eq . [ xref - equation-103193044 ] . the y - dependence of the class @xmath80 solutions is defined by the metric function @xmath209 . before we proceed to look at specific classes of solutions that satisfy theorem [ xref - theorem-103203143 ] , let us have a closer look at the differential equations that define this function , eqs . [ xref - equation-10322628 ] and [ xref - equation-10322635 ] . in the special case where both @xmath170 and @xmath91 , these equations both reduce , under the parametric constraint given by eq . [ xref - equation-5264516 ] , to the equation @xmath210 which has the solution @xmath211 where @xmath212 and @xmath213 are constants of integration . solutions . ] in all other cases , eqs . [ xref - equation-10322628 ] and [ xref - equation-10322635 ] are equivalent to the following two equations : @xmath214 where we have applied the constraints of theorem [ xref - theorem-103203143 ] and introduced the three parameters @xmath215 , @xmath212 and @xmath216 , defined as follows : @xmath217 the presence of these three parameters reflects that , in general , class @xmath80 solutions have three free parameters . one ( @xmath218 , or equivalently , @xmath68 ) is shared with class @xmath88 solutions . the second parameter , @xmath212 , defines the zero - point value of the effective scalar potential , @xmath106 $ ] . the last parameter , @xmath216 , is proportional to the cosmological constant , @xmath5 . equation [ xref - equation-528652 ] follows by elimination of the potential term , while eq . [ xref - equation-528726 ] follows by elimination of the second - order derivative . now , by taking the derivative of eq . [ xref - equation-528726 ] and applying the value of @xmath68 that is valid for @xmath136 , @xmath139 , we get eq . [ xref - equation-528652 ] . thus , any solution to eq . [ xref - equation-528652 ] is also a solution to eq . [ xref - equation-528726 ] . the general solution to these equations is @xmath219 where @xmath213 is an arbitrary constant of integration . notice that there are no real solution for @xmath110 with parameter values @xmath220 and @xmath221 . the three parameters @xmath215 , @xmath212 and @xmath222 is a convenient set of parameters . @xmath215 is a dimensionless parameter that takes any real value . let us express the parameters used by theorem [ xref - theorem-103203143 ] in terms of the new parameters . applying the parametric constraints stated by theorem [ xref - theorem-103203143 ] , the parameters @xmath56 and @xmath223 can conveniently be expressed in terms of the dimensionless parameter @xmath215 : @xmath224 where @xmath225 . this parametrization of @xmath56 and @xmath223 in terms of @xmath215 is valid for all solutions of classes @xmath88 and @xmath80 . similarly , @xmath226 and @xmath227 can be expressed in terms of @xmath215 as follows : @xmath228 finally , we can express the parameters @xmath157 and @xmath86 in terms of the new parameters @xmath215 , @xmath212 and @xmath222 : @xmath229 let us first look at the @xmath97class of solutions , i.e. the solutions with @xmath91 . there are two sub classes of solutions that we must treat separately : @xmath170 and @xmath230 . the simplest case is @xmath97 with vanishing scalar potential ; @xmath170 . in this case , the right hand sides of the @xmath128 and @xmath130 components of the einstein equations , equations [ xref - equation-103214451 ] and [ xref - equation-103193029 ] , both vanish . by comparing the left hand sides of these equations , we see that the constraint @xmath231 must be fulfilled in order for these two equations to not be in conflict and still yield solutions with @xmath232 the field equations reduce to the differential equation of eq . [ xref - equation-53185222 ] , which has the general solution @xmath233 . transforming to a new coordinate @xmath234 given by @xmath235 allows us to reexpress the solution on another form . let us summarize this set of solutions : @xmath236 the @xmath97 class of solutions with vanishing scalar field potential consists of a one - parameter family of solutions . the metric is given by [ xref - corollary-101818159 ] @xmath237 where the parameters @xmath56 , @xmath223 and @xmath238 are given by eqs . [ xref - equation-54222627 ] and [ xref - equation-55142955 ] with @xmath239 . @xmath68 and @xmath201 take the values @xmath68=@xmath240 and @xmath201=@xmath241 . @xmath212 is an arbitrary , positive constant of dimension @xmath242 . the massless scalar field is @xmath243where @xmath244and @xmath245 . next , let us consider the class of @xmath97solutions with non - vanishing scalar field potential ; @xmath172 . these solutions can be inferred directly from theorem [ xref - theorem-103203143 ] and eq . [ xref - equation-5465721 ] . in 4 dimensions and for static vacuum solutions , they are the solutions found by fonarev in @xcite . according to eq . [ xref - equation-5465721 ] , the metric function @xmath208 is in this case given by @xmath246 for the special case @xmath247 , the solution takes the form @xmath248 in the general case for @xmath249 , we do a coordinate transformation to a new coordinate @xmath234 defined by @xmath250 . we now get the metric on the form @xmath251 and the scalar field takes the form @xmath252 we see that the metric and scalar field for @xmath249 coincide with those for the case @xmath247 . thus , we can summarize this result as follows : @xmath253 the @xmath97class of solutions with non - vanishing scalar field potential is a two - parameter family of solutions given by the metric [ xref - corollary-109203724 ] @xmath254a scalar field on the form @xmath252and a scalar field potential @xmath155 = w_{0}e^{-2k \phi } , \]]where k=@xmath255 and @xmath256 , @xmath257 > 0 . corollary [ xref - corollary-101818159 ] in combination with corollary [ xref - corollary-109203724 ] generalizes the fonarev result @xcite to arbitrary dimension and for arbitrary orthogonal - symmetric seed metrics . we also see that the solutions of corollaries [ xref - corollary-101818159 ] and [ xref - corollary-109203724 ] have the same form , so corollary [ xref - corollary-109203724 ] can be trivially extended to also cover the solutions of corollary [ xref - corollary-101818159 ] by allowing parameter values @xmath258 and @xmath239 . now , turning our attention to the classes of solutions with @xmath138 , the @xmath95class of scalar field solutions , we begin by investigating the case of solutions with vanishing scalar field potential , which requires that @xmath170 . the constraints that must be satisfied in this case are listed in theorem [ xref - theorem-103203143 ] , and are equations [ xref - equation-10322143 ] , [ xref - equation-10321522 ] , [ xref - equation-103214959 ] and [ xref - equation-10982211 ] . the solution to these constraints is given by a discrete set of parametric values corresponding to the value @xmath259@xmath260 : @xmath261 where @xmath262 . according to eq . [ xref - equation-5465721 ] , the metric function @xmath110 takes the form @xmath263 , which requires @xmath264 . let us summarize this solution : @xmath265 the set of @xmath181 solutions with vanishing scalar field potential is a one - parameter family of solutions parametrized by the parameter @xmath222 with @xmath266 . the solutions correspond to a choice of cosmological constant in the seed solution satisfying @xmath267 . the other parameters are defined by equations [ xref - equation-54222627 ] - [ xref - equation-55165629][xref - equation-54222627]with @xmath268 , and consequently @xmath139 . the metric is given by [ xref - corollary-101818412 ] @xmath269with a scalar field @xmath270@xmath106 = 0 $ ] the final part of the @xmath180solution space to be analyzed is the @xmath95 class of solutions with non - vanishing scalar field potential . proceeding as before , we first identify the constraints to solve . in this case it is the full set of constraints , namely equations [ xref - equation-10322143 ] , [ xref - equation-10321522 ] , [ xref - equation-1098246 ] , [ xref - equation-103214959 ] , [ xref - equation-10982211 ] and [ xref - equation-103193044 ] . there are two unique solutions to these constraints : @xmath271 where @xmath262 , as before . these parameter values correspond to @xmath272 . the expression for @xmath110 is given by eq . [ xref - equation-5465721 ] . by transforming to new coordinates , we are able to simplify the expressions considerably . define a new coordinate @xmath273 by @xmath274 doing so , we are able to simplify the expressions considerably . let us summarize the result : @xmath275 the @xmath181class of solutions with non - vanishing scalar field potential is a two - parameter family of solutions given by the metric [ xref - corollary-1018183318 ] @xmath276a scalar field on the form @xmath277and a scalar field potential @xmath155 = w_{0}e^{-2k \phi } , \]]where @xmath278 and @xmath279 , corresponding to @xmath272 . @xmath212 and @xmath222 are free parameters of dimension @xmath242 . the @xmath95solutions must have @xmath280 and satisfy at least one of the conditions @xmath257 > 0 or @xmath281 < 0 . we will now seek a common form of all class @xmath88 and @xmath282 solutions that is capable of representing all solutions . in order to do so , we must introduce yet another set of parameters . we will continue to use the parameter @xmath218 used before , but define the new parameters @xmath283 and @xmath284 . writing @xmath257 and @xmath281 in terms of @xmath285 and @xmath286 gives @xmath287 and @xmath288 . it is now straight forward to prove our main result in terms of parameters @xmath215 , @xmath285 and @xmath286 , which we summarize in the form of a new theorem : given an orthogonal - symmetric vacuum solution , @xmath162g[v,@xmath163 , to the n dimensional einstein equations with an arbitrary cosmological constant @xmath86 ( equation [ xref - equation-92783053 ] ) , the complete set of class @xmath88 and @xmath80 solutions derived from @xmath41 has a space - time metric [ xref - theorem-101785048 ] @xmath289and a scalar field @xmath290with a scalar field potential @xmath155 = w_{0}e^{-2k \phi } , % \label{xref - equation-101785254}\]]where @xmath262 , @xmath218 and b are dimensionless parameters and q is a parameter of dimension inverse length . both b and q must be @xmath291 0 . the constants @xmath68 , @xmath201 , k and @xmath157 are given by @xmath292and the parameters q and b are related to @xmath86 as follows : @xmath293the main solution classes can be retrieved from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] as follows : * @xmath177 : the complete set of class @xmath177solutions is obtained from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] when @xmath294 and @xmath218 @xmath295 @xmath296 . * @xmath97 : the @xmath97 solutions are obtained from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] when @xmath297 , @xmath298 and @xmath218 @xmath295 @xmath299 . * @xmath95 : the @xmath181solutions are obtained from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] when @xmath300 , @xmath298 and @xmath301 . if we first consider the case of the empty @xmath178 solution class of corollary [ xref - corollary-1018173233 ] , we see that the fact that there are no solutions of this class is reflected by equations [ xref - equation-1018173355 ] and [ xref - equation-101817345 ] , because both @xmath172 and @xmath136 require @xmath302 . from the metric in equation [ xref - equation-101817047 ] we see that this means that a class @xmath80 solution is generated , thus there are no solutions of class @xmath92 . next , corollary [ xref - corollary-104611 ] follows directly from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] when setting @xmath303 . corollary [ xref - corollary-101818159 ] follows directly from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] when we let @xmath304 , @xmath305 and[xref - corollary-101818159]@xmath297[xref - corollary-101818159 ] similarly , corollary [ xref - corollary-109203724 ] can be inferred from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] by letting @xmath304 and setting @xmath297 . corollary [ xref - corollary-101818412 ] follows from equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] by letting @xmath306 , @xmath307 , @xmath308 and @xmath309 . finally , the results of corollary [ xref - corollary-1018183318 ] can be brought to the form of equations [ xref - equation-101817047 ] and [ xref - equation-10181714 ] by transforming to a new coordinate @xmath234 defined by @xmath310 the completeness of theorem [ xref - theorem-101785048 ] follows from the way we derived it : a general form of the potential was investigated , from which we derived the potential of equation [ xref - equation-101785254 ] . furthermore , we did an exhaustive analyzis of all options for satisfying the constraints set up by theorem [ xref - theorem-103203143 ] , which represent all class @xmath80 and @xmath88 solutions . no solutions within the @xmath95 class of solutions have been published before , so let us have a closer look at a particular solution of this class in 4 dimensions . as the class requires seed solutions with non - zero cosmological constant , let us use the ( anti- ) desitter solution in 4 dimensions as our seed solution . take @xmath311 as a non - zero cosmological constant . the @xmath0 metric can be cast into the following static form : @xmath312 define @xmath313 . applying corollary [ xref - corollary-1018183318 ] and defining a new coordinate @xmath314 allows us to cast the solution metric into the following form : @xmath315 where @xmath316 for @xmath317 , while @xmath318 when @xmath319 . likewise , @xmath320 if @xmath317 , while @xmath321 when @xmath322 . in the case of positive @xmath5 , the spatial slices of this geometry are conformally equivalent to the hypersphere @xmath323 and the solution topology is @xmath324 , while for negative @xmath5 , the spatial slices are conformally equivalent to the hyperboloid @xmath325 and the solution topology is @xmath326 . the scalar curvature is @xmath327 the scalar field takes the unique form @xmath328 which is independent of parameters @xmath212 and @xmath281 . the metric of eq . [ xref - equation-5106438 ] is singular for time @xmath329 , while the ricci scalar of eq . [ xref - equation-51064940 ] is regular . by transforming the riemann tensor to an orthonormal basis , we find that @xmath330 is just a coordinate singularity . the geometry has an initial singularity at @xmath331 , and there is a curvature singularity at @xmath332 for @xmath333 . as @xmath334 , the scalar field @xmath335 . the parameters @xmath212 and @xmath222 evaluate to @xmath336 and @xmath337 , so @xmath338 . we will now transform the de sitter - based solution we generated in the previous section to the conformal frame , using the generalized bekenstein transformation for arbitrary scalar potential @xcite . the generalized bekenstein transformation is described in theorem [ xref - theorem-510234826 ] of appendix [ xref - appendixsection-510235039 ] . applying it to the solution of eqs . [ xref - equation-5106438 ] and [ xref - equation-511004 ] for @xmath317 gives a metric of the transformed solution : @xmath339 where @xmath340 , and we have introduced a new time coordinate @xmath341 . the conformally coupled scalar field @xmath342 is @xmath343 with @xmath344 . the scalar potential of this solution is @xmath345 = { w_{0 } ( 1-\zeta ^{2}\phi ^{2 } ) } ^{2}e^{-2\zeta \phi } = w_{0 } ( 1-\zeta ^{2}\phi ^{2 } ) { \left ( 1-\zeta \phi \right ) } ^{2}.\ ] ] the ricci scalar of this geometry is @xmath346 thus , there is a curvature singularity where @xmath347 . thus , for @xmath348 , the geometry is regular everywhere . by transforming the riemann tensor to an orthonormal basis , we find that the solution is free of other singularities . introduce constants @xmath349 and @xmath350 . then define a new time coordinate @xmath351 by @xmath352 , where @xmath353 . we are now able to cast the metric into the following form : @xmath354 where @xmath355 is a dimensionless constant , @xmath356 , and @xmath357 is a time - dependent scale factor ; @xmath358 . the cosmological time scale is @xmath359 . similarly , @xmath360 is a cosmological length scale that is identical to the length scale of the @xmath361 seed metric , eq . [ xref - equation-526103121 ] . we can then write the ratio @xmath362 in terms of @xmath363 and @xmath364 : @xmath365 . notice that @xmath366 is proporional to @xmath367 , so @xmath355 is defined by the relative size of the cosmological constant over the strength of the scalar potential . in the limit @xmath368 , the geometry of eq . [ xref - equation-513141035 ] is spatially homogenous . we can therefore refer to @xmath355 as the _ inhomogeneity parameter _ of the spacetime . a cosmological constant that is weak relative to the strength of the scalar potential yields a geometry that is highly homogeneous ( low @xmath355 ) , while a strong cosmological constant yields a highly inhomogenous cosmology ( high @xmath355 ) . the scale factor @xmath357 is @xmath21 for @xmath369 . this solution is clearly a spatially finite , inhomogeneous , inflationary cosmology with a time scale set by @xmath364 and a size of the spatial sections governed by the length scale @xmath370 and the scale factor @xmath371 . corresponding to times @xmath372 , this universe has an initial curvature singularity that vanishes at time @xmath373 and a final curvature singularity that starts to emerge from @xmath374 at time @xmath375 . from eq . [ xref - equation-5129038 ] we see that the geometry has positive curvature everywhere . it has a finite volume of @xmath376 thus , the size of the cosmology is defined by the time constant @xmath359 , the scale factor @xmath357 , as well as the inhomogeneity @xmath355 . a weak scalar potential implies a long time scale and a small cosmological constant implies a cosmology of large size . furthermore , if @xmath377 , the time span between the initial and final singularities are much larger than the cosmological time scale : @xmath378 , the geometry is nearly static because , at @xmath369 , @xmath379 and @xmath380 : @xmath381 this means that the size of the cosmology is nearly constant at times @xmath382 ; in fact @xmath379 for @xmath369 , but the geometry is still accelerating . at late times , @xmath383 , the cosmology approaches power - law inflation : @xmath384 now , if we define a new time coordinate @xmath385 , the metric of eq . [ xref - equation-513141035 ] takes the form @xmath386 where the conformal time @xmath387 is defined in the interval @xmath388 . thus , the solution geometry @xmath389 is conformally isometric to the einstein static universe @xmath390 , and maps onto the @xmath391 section of the einstein cylinder . then , @xmath392 as @xmath393 . the conformal boundary of the spacetime consists of the two timelike conformal infinities @xmath394 at @xmath395 , respectively . all timelike geodesics emerge from @xmath396 and end at @xmath397 . if we analyze null geodesics on @xmath398 , we find that null geodesics in the @xmath399 direction satisfy @xmath400 , where @xmath401 . thus , due to rotational symmetry of @xmath323 , photons can only travel an angular distance @xmath402 in a finite amount of time in any spatial direction of @xmath398 . subsequently , for a stationary observer at an arbitrary spatial location , there is a cosmological event horizon at an angular distance @xmath2 away from the observer , implying that the observable universe is half the total universe . if we transform the cosmological horizon back to the physical solution geometry of @xmath389 , the horizon will be at the same spatial coordinates on @xmath389 as on @xmath398 , but the geometry of the horizon will be changed by the conformal factor of eq . [ xref - equation-51318159 ] . still , the observable universe extends an angular distance of @xmath403 away from the stationary observer . figure [ xref - figurecaption-5307522 ] shows a penrose - carter diagram for this spacetime . we see from this diagram that @xmath399-directed null geodesics , with the exception of the points on the @xmath404 surface , either emerge from the initial singularity or terminate at the final singularity . furthermore , there are timelike geodesics that do not intersect any of the singularities . we can also see that in the region @xmath405 , all @xmath399-directed geodesics that are either null or time - like intersect the final singularity . furthermore , we can infer from the penrose - carter diagram that for all solutions with @xmath406 , the null surface @xmath407 constitutes an event horizon , because no null or time - like geodesics with @xmath408 will ever enter the @xmath409 region . the solution is regular at the @xmath410 surface . from the metric of eq . [ xref - equation-51318159 ] , we see that the initial and final singularities are in fact point singularities , because the 2 dimensional surfaces @xmath411 have vanishing proper surface area as @xmath412 . we can conclude that the @xmath406 solutions are solutions in which a scalar black hole emerges . the event horizon emerges at conformal time @xmath413 , whereas the curvature singularity emerges later , at conformal time @xmath414 . -directed null geodesics . initial and final singularities are plotted for @xmath415 ( long dash ) , @xmath416 ( short dash ) , @xmath417 ( continuous ) . the black hole event horizon is the @xmath418 curve , and the gray region @xmath419 is the black hole region for all solutions with @xmath420 . ] in the terms of the conformal time @xmath387 , the scalar field takes the following form : @xmath421 we see from eq . [ xref - equation-514234249 ] that the scalar field blows up at the singularity , but is regular everywhere else , including at the event horizon . let us look at energy conditions . the weak and strong energy conditions can be expressed in terms of the energy density @xmath226 and the principal pressures @xmath422 evaluated in the rest frame of the spacetime fluid @xcite . the weak energy condition requires @xmath423 and @xmath424 , while the strong energy condition requires @xmath424 and @xmath425 . using the einstein equations , we may evaluate @xmath226 and @xmath422 from the einstein tensor : @xmath426 for the metric of eq . [ xref - equation-51318159 ] , this gives @xmath427 from eq . [ xref - equation-514215635 ] it follows immediately that the energy density is positive definite only for small @xmath355 ( @xmath428 ) and large c ( @xmath429 ) . furthermore , eqs . [ xref - equation-514215836 ] and [ xref - equation-514215856 ] yield @xmath424 on all of @xmath389 . thus , the scalar field satisifes the weak energy condition on all of @xmath389 for @xmath428 or @xmath429 . furthermore , @xmath430 @xmath431 may take both positive and negative values on @xmath432 , so the strong energy condition is generally violated on @xmath389 . finally , let us have a look at the equation of state parameter @xmath433 . defining @xmath434 , eqs . [ xref - equation-514215836 ] and [ xref - equation-514215856 ] allow us to write @xmath435 @xmath436 when @xmath437 . thus , the weak energy condition is satisfied when @xmath437 . we also see that @xmath438 as @xmath439 . this happens when the cosmological constant is much weaker than the strength of the scalar potential . in that case , we saw from eq . [ xref - equation-514234249 ] that the scalar field is nearly constant and approaches the planckian value @xmath440 . ( the spatial region @xmath441 has been chopped off ) . each plot shows a fixed - time projection of the spatial geometry for @xmath442 . top left : @xmath413 . top right : @xmath443 . bottom left : @xmath444 . bottom right : @xmath445 . the black hole event horizon is shown as a black curve . the top two plots show the spatial geometry before the singularity has formed . the bottom two plots show the spatial geometry after the singularity has formed . , width=566 ] figure [ xref - figurecaption-51502747 ] shows four different snapshots of the @xmath442 projections of the spatial geometry of the solution . in the second plot , the spatial geometry is still regular , but it has an event horizon , shown in black . in the next plot , the singularity has formed , and a black hole has been created . the protuberance in the middle of the two plots at the bottom is a spatial region that lies beyond the singularity , and is shown in the penrose - carter diagram of figure [ xref - figurecaption-5307522 ] as the spatial region to the right of the singularity . as a function of conformal time z for the two angular positions @xmath399=0 and @xmath399=@xmath446/3 . the curves are for @xmath447(continuous ) , @xmath448 ( short dash ) and @xmath449(long dash ) . ] as a function of conformal time @xmath387 for the two angular positions @xmath399=0 and @xmath399=@xmath446/3 . the curves are for @xmath447(continuous ) , @xmath448 ( short dash ) and @xmath449(long dash ) . ] figure [ xref - figurecaption-51665528 ] plots the scalar field as a function of conformal time @xmath387 and angular coordinate @xmath450 for three different values of the coordinate @xmath355 . figure [ xref - figurecaption-5150107 ] plots the equation of state parameter @xmath451 as a function of conformal time @xmath387 and angular coordinate @xmath452 for three different values of the constant @xmath355 . in this paper , we generalized two previously known solution generation techniques for generating minimally coupled einstein - scalar field solutions in 4 dimensions ( the buchdahl @xcite and fonarev @xcite transformations ) . two generalizations were made : i ) the transformation was generalized to arbitrary dimension , and ii ) the new transformation allows vacuum solutions with non - zero cosmological constant as seed . thus , we are now able to generate minimally coupled einstein - scalar field solutions from vacuum solutions with arbitrary cosmological constant in arbitrary dimension . the only requirement to a seed solution is that it posesses a hypersurface - orthogonal killing vector field . we introduced a labeling scheme that labels all solutions in terms of whether the target solution is invariant with respect to translations along the killing vector field ( class @xmath77 solutions ) or not ( class @xmath79 solutions ) . furthermore , we labeled the target solution with dimension @xmath3 , as well as whether the seed solution has non - zero cosmological constant @xmath453 or not ( @xmath454 ) . the solutions cover new , unknown solutions @xmath455 as well as previously known solutions , such as the solutions of buchdahl @xmath456 , fonarev ( @xmath457 ) , feinstein - kunze - vzquez - mozo ( @xmath458 ) and xanthopoulos & zannias @xcite ( static , spherically symmetric @xmath197 ) . the generalization of the buchdahl transformation to arbtirary dimension is new @xmath459 . using the generalized buchdahl transformation , we are able to recapture the extra - dimensional static and spherically symmetric solutions of xanthopoulos and zannias as a special case . the generalization that allows us to use seed solutions with non - zero cosmological constant uncovers a new class of einstein - scalar field solutions that has previously not been studied . we apply our solution generation technique in order to study one of the familiy of solutions , generating einstein - scalar field solutions from the @xmath0 vacuum solutions . the resulting einstein - scalar field solution that comes from transforming the de sitter vacuum is a two - parameter family of 4 dimensional , inhomogenous , expanding cosmological solutions with @xmath324 topology and exponential scalar potential parametrized by two parameters : the strength of the scalar potential , @xmath157 , and the ( positive ) cosmological constant @xmath5 of the seed solution . by transforming a solution of this kind to the conformal frame using the generalized bekenstein transformation , we find a spatially finite , expanding and accelerating cosmological solution that is conformally isometric to the einstein universe @xmath1 . the solution can be parametrized by a cosmologial time scale @xmath364 , defined by the strength of the scalar potential , and the _ inhomogeneity parameter _ @xmath355 , defined by the ratio @xmath460 . @xmath364 defines the scale of the solution , while @xmath355 defines the degree of spatial inhomogeneity . low @xmath355 implies that the solution is highly homogeneous . we study null geodesics and find that for any observer , the solution has a cosmological event horizon at an angular distance of @xmath2 away from the observer . the solution has an initial singularity , which is a point singularity that vanishes at early times . the solution is then free of singularities until a final singularity emerges at late times as a new point singularity . there are timelike and null geodesics that do not intersect any of the singularities . for @xmath406 , the late time singularity is hidden inside an event horizon . this family of solutions therefore has the natural interpretation of being expanding cosmologies in which a scalar black hole is formed at late times . the energy density is positive definite for small @xmath355 ( @xmath461 ) and large @xmath355 ( @xmath429 ) . the conformally coupled scalar field satisfies the weak energy condition as long as the energy density is positive , while the strong energy condition is generally violated . i would like to thank professor finn ravndal at the university of oslo for very valuable discussions while preparing this paper . here , we will briefly summarize how certain covariant quantities transform under conformal transformations . for a reference , see e.g. @xcite . let @xmath113 and g be the metrics of two @xmath462-dimensional space - time geometries that are related by a conformal transformation @xmath463 as follows : @xmath464 let @xmath465 be an arbitrary covariant @xmath462-vector . covariant derivatives transform as follows under a conformal transformation @xmath466 : @xmath467 where @xmath468 is the covariant derivative of the metric @xmath113 . recognizing that @xmath469 @xmath470 , we may use equation [ xref - equation-21662857 ] to compute how second - order covariant derivatives transform under conformal transformations : @xmath471 the ricci tensor @xmath120transforms as follows under a conformal transformation @xmath472 : @xmath473 = { } { } r_{\mu \nu } [ g ] + \frac{2\left ( m-2\right ) } { \omega ^{2}}\nabla _ { \mu } \omega \nabla _ { \nu } \omega -\frac{\left ( m-3\right ) } { \omega ^{2}}g_{\mu \nu } \nabla ^{\alpha } \omega \nabla _ { \alpha } \omega -\frac{\left ( m-2\right ) } { \omega } \nabla _ { \mu } \nabla _ { \nu } \omega -\frac{1}{\omega } g_{\mu \nu } \box \omega % \label{xref - equation-215174554}\ ] ] writing equation [ xref - equation-215174554 ] in terms of @xmath474 gives : @xmath473 = { } { } r_{\mu \nu } [ g ] -\left ( m-2\right ) \nabla _ { \mu } \nabla _ { \nu } \omega + \left ( m-2\right ) \nabla _ { \mu } \omega \nabla _ { \nu } \omega -g_{\mu \nu } ( \box \omega + \left ( m-2\right ) \nabla ^{\alpha } \omega \nabla _ { \alpha } \omega ) % \label{xref - equation-21664714}\ ] ] from equation [ xref - equation-215174554 ] we can compute the ricci scalar of the transformed metric : @xmath475 \equiv { \overline{g}}^{\mu \nu } { \overline{r}}_{\mu \nu } = \omega ^{-2 } ( r [ g ] -\left ( m-1\right ) \left ( \frac{2}{\omega } \box \omega + \left ( m-4\right ) \nabla ^{\alpha } \omega \nabla _ { \alpha } \omega \right ) ) \ ] ] it then follows that the einstein tensor @xmath476 = r_{\mu \nu } [ \overline{g } ] -\frac{1}{2}{\overline{g}}_{\mu \nu } r [ \overline{g}]\ ] ] transforms as follows : @xmath477 + \frac{\left ( m-4\right ) } { 6}g_{\mu \nu } \nabla ^{\alpha } \omega \nabla _ { \alpha } \omega \right ) , % \label{xref - equation-21610338}\ ] ] where @xmath478 \equiv \nabla _ { \mu } \phi \nabla _ { \nu } \phi -\frac{1}{2}g_{\mu \nu } \nabla ^{\alpha } \phi \nabla _ { \alpha } \phi \\\theta _ { \mu \nu } [ g,\phi , \chi ] \equiv s_{\mu \nu } [ g,\phi ] -\chi ^{2}\nabla _ { \mu } \nabla _ { \nu } \phi ^{2}+\chi ^{2}g_{\mu \nu } \box \phi ^{2}.\end{gathered}\ ] ] here , g denotes an arbitrary metric , @xmath47 denotes an arbitrary scalar function , and @xmath281 is an arbitrary constant . the non - minimally coupled einstein - scalar field action in dimension @xmath3 is @xmath479 \equiv \int d^{n}x\sqrt{-g}\left ( \frac{1}{2\kappa } \left ( r-2\lambda \right ) -\frac{1}{2}\nabla ^{\alpha } \phi \nabla _ { \alpha } \phi -\frac{1}{2}\xi ^{2}r \phi ^{2}-v [ \phi ] \right ) , \ ] ] where @xmath480 is the dimensionless scalar - gravity coupling constant . we have conformal coupling when @xmath481 . define @xmath482 . when extremalizing this action , we obtain the conformally coupled einstein - scalar field equations : @xmath483 = \frac{\kappa } { \left ( 1-\zeta ^{2}\phi ^{2}\right ) } \left ( \theta _ { \mu \nu } [ g,\phi , \xi ] - g_{\mu \nu } w [ \phi ] \right ) % \label{xref - equation-510233227 } \\\box \phi -\xi ^{2}r \phi -\frac{d w [ \phi ] } { d \phi } = 0,% \label{xref - equation-510233249}\end{gathered}\ ] ] let @xmath486 be an @xmath3 dimensional solution to the minimally coupled einstein - scalar field equations , and let @xmath487 be an @xmath3 dimensional solution to the non - minimally coupled einstein - scalar field equations with arbitrary scalar - gravity coupling @xmath480 . a _ bekenstein transformation _ @xmath488 is defined as a conformal transformation @xmath463 mapping the metric @xmath13 into @xmath489 and a function @xmath490 mapping the scalar field @xmath342 into @xmath47 : @xmath491\end{gathered}\ ] ] _ for every minimally coupled einstein - scalar field solution , there are two , and only two , bekenstein transformations that relate the minimally coupled solution to two einstein - conformal scalar field solutions . if _ @xmath486_is an _ @xmath3_dimensional solution to the einstein - scalar field equations with a minimally coupled scalar field _ @xmath47_and a scalar field potential _ @xmath492 , _ there are two independent solutions _ @xmath487 , labeled a and b , _ to the einstein - conformal scalar field equations , eqs . _ [ xref - equation-510233227]_and _ [ xref - equation-510233249 ] _ , given by_[xref - theorem-510234826 ] @xmath493 = \begin{cases } \omega ^{n}\overline{w } [ \phi [ \phi ] ] = \|1-\zeta ^{2}\phi ^{2}\|^{\frac{n}{n-2}}\overline{w } [ \phi [ \phi ] ] & \left ( a\right ) \\ -\omega ^{n}\overline{w } [ \phi [ \phi ] ] = -\|1-\zeta ^{2}\phi ^{2}\|^{\frac{n}{n-2}}\overline{w } [ \phi [ \phi ] ] & \left ( b\right ) \\ \end{cases}\end{gathered}\ ] ] 000 d. n. spergel and . et . al ( 2006 ) . wilkinson microwave anisotropy probe ( wmap ) three year results : implications for cosmology . arxiv : astro - ph/0603449 [ spergel:2006hy ] b. ratra and p. j. peebles , phys . , * d37 * , 3406 , ( 1988).[ratra:1987rm ] r. r. caldwell , r. dave and p. j. steinhardt , phys . * 80 * , 1582 , ( 1998).[caldwell:1997ii ] l. wang , r. r. caldwell , j. p. ostriker and p. j. steinhardt , astrophys . j. , * 530 * , 17 , ( 2000).[wang:1999fa ] p. g. ferreira and m. joyce , phys . rev . , * d58 * , 023503 , ( 1998).[ferreira:1997hj ] j. polchinski , _ string theory volume ii . superstring theory and beyond _ ( cambridge univ . press , cambridge , 1998).[polchinski-1998 ] o. bertolami , j. paramos and s. g. turyshev ( 2006 ) . general theory of relativity : will it survive the next decade ? . preprint arxiv : gr - qc/0602016 [ bertolami - paramos - turyshev-2006 ] i. k. wehus and f. ravndal ( 2006 ) . gravity coupled to a scalar field in extra dimensions . preprint arxiv : gr - qc/0610048 [ wehus - ravndal-2006 ] h. a. buchdahl , phys . , * 115 * , 1325 , ( 1959).[buchdahl:1959nk ] o. bergmann and r. leipnik , phys . , * 107 * , 1157 , ( 1957).[bergmann - leipnik-1957 ] a. i. janis , e. t. newman and j. winicour , phys . * 20 * , 878 , ( 1968).[janis - newman - winicour-1968 ] a. i. janis , d. c. robinson and j. winicour , phys . rev . , * 186 * , 1729 , ( 1969).[janis-1969 ] b. c. xanthopoulos and t. zannias , phys . , * d40 * , 2564 , ( 1989).[xanthopoulos:1989 kb ] c. callan , s. coleman and r. jackiw , ann . * 59 * , 42 , ( 1970).[callan-1970 ] j. d. bekenstein , ann . physics , * 82 * , 535 , ( 1974).[bekenstein-1974 ] j. froyland , phys . , * d25 * , 1470 , ( 1982).[froyland:1982 yd ] k. maeda , phys . , * d39 * , 3159 , ( 1989).[maeda:1988ab ] v. husain , e. a. martinez and d. nunez , phys . , * d50 * , 3783 , ( 1994).[husain:1994uj ] o. a. fonarev , class . quant . grav . , * 12 * , 1739 , ( 1995).[fonarev:1994xq ] a. feinstein , k. e. kunze and m. a. vazquez - mozo , phys . , * d64 * , 084015 , ( 2001 ) . arxiv : hep - th/0105182 [ feinstein:2001xs ] r. m. wald , _ general relativity _ ( the university of chicago press , chicago and london , 1984).[wald:1984-general ]	this paper generalizes two previously known techniques for generating minimally coupled einstein - scalar field solutions in 4 dimensions ; the buchdahl and fonarev transformations . two generalizations are made : i ) the transformation is generalized to arbitrary dimension , and ii ) the new transformation allows vacuum solutions with non - zero cosmological constant as seed . thus , by applying this solution generation technique , minimally coupled einstein - scalar field solutions can be generated from vacuum solutions with arbitrary cosmological constant in arbitrary dimension . the only requirement to a seed solution is that it posesses a hypersurface - orthogonal killing vector field . the generalization that allows us to use seed solutions with arbitrary cosmological constant uncovers a new class of einstein - scalar field solutions that has previously not been studied . we apply the new solution transformation to the @xmath0 vacuum solution . transforming the resulting einstein - scalar field solution to the conformal frame , a two - parameter family of spatially finite , expanding and accelerating cosmological solutions are found that are conformally isometric to the einstein static universe @xmath1 . we study null geodesics and find that for any observer , the solution has a cosmological horizon at an angular distance of @xmath2 away from the observer . a subset of these solutions are studied in particular : a solution of this kind has an initial point singularity that vanishes at early times as well as another point singularity that emerges at late times . the solution is non - singular in between those events . the late time singularity is hidden behind an event horizon , and these solutions can therefore be naturally interpreted as expanding cosmologies in which a scalar black hole is formed at late times . the energy density is positive definite only in parts of the parameter space . the conformally coupled scalar field satisfies the weak energy condition as long as the energy density is positive , while the strong energy condition is generally violated . [ i1 ]
spin - wave generation and manipulation have already been demonstrated in many ways.@xcite they are of great importance in magnetism - based spin - wave logic and xnor as well as nand gates have recently been implemented.@xcite to reduce the size of potential devices , effective spin - wave pumping mechanisms are needed in ferromagnetic metallic films owning micron to nanometer characteristic magnetic length scales . here we present a pumping mechanism using fs - laser excitation that allows for a large precession angle ( @xmath1 ) after local excitation . on the other hand , the relevant length scales can be artificially selected by the design of magnonic crystals , i.e. periodically micro- and nano - structured magnetic materials , with only very selected dynamic eigen states . for large precession angles the equations of motion are intrinsically non - linear . this can be used to pump energy into a certain spin - wave mode by non - linear interaction and becomes evident for example in spin - wave nano - oscillators.@xcite given a spatial separation of two spin - torque oscillators and hence , an overlap in their spin - wave power when excited by a direct current , one finds a locking of frequencies if the frequency mismatch is below a critical value . driven to an extreme , a steady transfer of energy from higher to lower energy modes results in bose - einstein condensation observed at room temperature.@xcite under an external field of @xmath2 . a large angle of precession as high as @xmath3 is observed . inset : schematic representation of spin - wave modes in a continuous ferromagnetic film of thickness @xmath4 . local amplitude and phase of the dynamic magnetization @xmath5 are illustrated . shape anisotropy accounts for mainly in - plane alignment of the magnetization @xmath6 . ] in the following we focus on dipolar and exchange spin - wave modes in continuous films excited by intense laser pulses , where the nature of the excited spin - wave modes is tuned by the external magnetic field . the excitation by femtosecond laser pulses has major differences as compared to methods that work in thermal equilibrium . these are for example brillouin light scattering ( bls),@xcite conventional ferromagnetic resonance ( fmr ) , strip - line based techniques like vector network analyzer ( vna)@xcite or spatio - temporal kerr effect based fmr used to study spin - wave modes.@xcite our technique works in space and time - resolved manner and the detection process owns no @xmath7-selectivity a priori . yet , the strongest difference to resonant techniques is the broadband excitation in the spin system,@xcite by which all magnetic modes are excited . however , only resonant modes contribute to the signal coherently and can be detected while others appear as non - coherent background . a further advantage is the contact - free , very local excitation and detection which simplifies the testing of various samples and structured media . we make the very unusual observation , that below a critical magnetic field two dominant spin - wave modes are excited , whereas above the critical field , they merge into one single mode . we shall also see , that intrinsic to the experimental method used , surface modes are preferentially excited if the sample thickness considerably exceeds the penetration depth of the laser field . in general , magnetic precessional motion can have dipolar and exchange spin - wave character and can be described as follows . starting from the free magnetic energy density , the landau - lifshitz equation of motion for the magnetization vector @xmath8 yields a theoretical dispersion describing the uniform precession ( also referred to as the kittel mode ) in the macrospin approximation@xcite @xmath9 where @xmath10 is the saturation magnetization ( being @xmath11 for nickel ) , the only free parameter @xmath12 accounts for the effective anisotropy in the out - of - plane direction , and @xmath13 is the component of the external field projected to the sample plane . damon and eshbach solved the equation of motion accounting for maxwell s equations in the magnetostatic limit and adequate boundary conditions yielding the dispersion for long wavelength surface modes , so called damon - eshbach ( de ) modes : @xmath14.\ ] ] here , @xmath4 is the film thickness and @xmath15 denotes the wave vector of the magnetostatic surface wave . an analogous approach in ref . [ ] yields dispersion characteristics taking into account exchange interaction as well as anisotropy . in thin magnetic layers one finds a quantization of the wave vector perpendicular to the sample plane and in analogy to eq . one can give a dispersion describing the exchange - dominated perpendicular standing spin waves ( pssw ) : @xmath16 therein , @xmath17 is the exchange constant and @xmath18 is the quantized wave vector in out - of - plane ( i.e. @xmath19)-direction attributed to a given order @xmath20 of the pssw . a schematic of the different magnetic modes is given in the inset of where the kittel mode is not explicitly drawn , as both pssw and de modes result in the uniform precession in the limit @xmath21 ( i.e. @xmath22 ) and @xmath23 , respectively . the samples consist of polycrystalline nickel films with thicknesses of up to @xmath24 . the thickness dependence of the occurring modes is measured on a wedge - shaped sample to perform all measurements on one specimen . it was prepared by electron beam evaporation in ultra high vacuum under a base pressure of @xmath25 . a linearly moving shutter was used to produce a nickel wedge with thickness @xmath26 on a si(100 ) substrate and in order to prevent oxidation , the ferromagnetic layer was capped with @xmath27 of copper . for the experiments we use an approach in the time domain , i.e. an all - optical pump - probe setup as previously described in reference [ ] . an ultrashort laser pulse with @xmath28 duration and @xmath29 central wavelength excites the saturated sample and at a variable time delay @xmath30 a second laser pulse with the same characteristics but @xmath31 of the intensity probes the time - resolved magneto - optical kerr effect ( trmoke).@xcite therewith , spin dynamics in the femtosecond regime can be detected , where electronic excitations lead to a demagnetization and subsequent relaxation into the initial state via scattering of high energy to low energy spin waves.@xcite the processes that initially quench the magnetization in the first few picoseconds are subject to ongoing discussions,@xcite but can be modeled quite successfully in some detail.@xcite here in focus are oscillations on a timescale from @xmath32 to @xmath33 . we shall see , that all modes corresponding to equations to can be optically excited in a thin ferromagnetic film . the responsible mechanism can be thought as an effective field pulse which stems from the heat - induced change of the sample s anisotropy and local exchange field upon absorption of the pump pulse.@xcite therefore , to promote magnetic precession the external field was tilted by @xmath34 out - of - plane thus giving rise to an angle between sample plane and the effective internal field @xmath35 . for an external field of @xmath36 , which is the maximum used in the experiments , we calculated a rotation of @xmath37 of @xmath38 out - of - plane due to the zeeman term in balance with the shape anisotropy ( see also ) . the angle is rather small and thus , its influence can be neglected for simplicity . hence , in the analytical expression of the spin - waves modes we assume @xmath39 and @xmath40 to be in - plane . for varying thickness @xmath4 . below @xmath41 only the uniform precession ( white points , black line ) is observed , which does not change in frequency for increasing @xmath4 . perpendicular standing spin waves with @xmath42 can also be identified ( gray ) and fit well to the expected @xmath43 behavior.@xcite pssw of second order are not observed ( dashed gray line ) , instead damon - eshbach surface waves ( blue ) are excited . the blue dashed line represents the de dispersion with @xmath44 and reveals that the de wave vector is not constant in our experiment ( the solid blue line is a guide to the eye ) . ] in order to obtain the precession frequencies from the time - dependent moke spectra , the incoherent background owing to phonons and magnons is subtracted . this procedure is demonstrated in with a single measurement recorded at @xmath45 and @xmath46 . after subtraction of the background ( black data points in ) , fourier transformation of the @xmath47 curves yields peaks in the oscillations power spectra . for a given thickness @xmath4 , the external field is varied between @xmath48 and @xmath49 , resulting in a change of amplitude and frequency of the oscillation . the position of the corresponding peaks in the frequency domain is determined for each such set of measurements and plotted versus @xmath50 ( for details of the analysis refer to the supplementary data ) . as calculated from the time - resolved moke measurements at @xmath51 , @xmath52 and @xmath53 , respectively . solid and dashed lines represent lorentz - peaks which where fitted to the data . the expected linewidth in nickel is larger than @xmath54 and as a result , possible mode locking leads to a single peak at @xmath55 if the overlap indicated by the hatched area is increased . additionally , the mutual intensity is strongly increased . in the inset , the dependence of the damon - eshbach wave vector on the thickness is given , where the solid gray line is a guide to the eye ( see text also ) . ] shown in are power spectra recorded on different nickel thicknesses , revealing up to three precessional modes of different origin . in the upper row the fourier power spectra for a given thickness are plotted in a color map as calculated from the @xmath47 curves ( the fourier transformation of the data from is marked with a vertical gray line ) . in the bottom row all spectra have been normalized by their respective maximal fft power to give a better overview of mode evolution . the data points represent the peak positions and included as solid lines are the fitted theoretical frequency dispersions - of the various modes , which have been used to attribute the branches to the kittel mode , the de surface mode as well as the pssw mode with @xmath42 . analysis of the thickness dependent occurance of the different magnetic modes yields the results depicted in . for a given external field of @xmath56 the observed frequencies have been extracted and the aforementioned identification of kittel- , pssw- , and de - mode is confirmed . both the kittel mode ( black ) and the pssw ( gray ) display the expected evolution with the tickness : while the frequency of uniform precession @xmath57 remains unchanged , the frequency of the pssw reduces . being proportional to @xmath43 it approaches @xmath57 in accordance to literature.@xcite the de modes ( blue lines and points in ) have been predominantly studied with brillouin light scattering ( bls)@xcite . we find from our experiment that they are first excited at a nickel thickness of @xmath58 . there , they are only present at low external fields , but for thicker nickel layers these modes dominate the spin - wave dynamics over a wider field range up to @xmath59 . this is due to the excitation mechanism in combination with the relevant length scales in our experiment : sample thickness @xmath4 and optical penetration depth @xmath60 ( i.e. penetration depth of the electrical field at optical frequencies which evaluates to @xmath61 , where @xmath62 is the complex refractivity index ) . calculations show that our experiment is sensitive to magnetic modes in the top @xmath63 of the ferromagnet , corresponding details will be published elsewhere.@xcite the magnetic excitation depth is governed by the penetration of the laser field and energy is deposited in form of an initial disarrangement of spins.@xcite this energy will very effectively excite eigen modes with similar spatial profile and thus , by choosing a thickness @xmath64 , the resulting strong asymmetry will make de surface modes the favored relaxation channel . consequently , the appearance of the de modes is based on the fact that they own an amplitude profile with the maximum at the surface , decaying exponentially into the film . herein lies a strong analogy to surface acoustic waves excited with laser pulses , which can be used to determine elastic properties of surface layers.@xcite assuming a propagation direction perpendicular to the magnetization ( @xmath65 ) the dispersion can be used to determine the wave vector @xmath66 . in the inset of the respective quantitative analysis of the damon - eshbach modes is given . from the fits one obtains wave vectors in the inverse micrometer range , corresponding to wave lengths of about @xmath67 . we find that the product @xmath68 entering eq . is not constant but that the relation @xmath69 holds with @xmath70 ( solid gray line in the inset of ) . the divergence of @xmath66 around that value supports our earlier interpretation . an asymmetry between film thickness and optical excitation depth is needed to give rise to dipolar spin waves in contrast to exchange - dominated spin waves with rather high @xmath7 . in other words , the parameter @xmath71 can be considered an onset thickness above which the dipolar interaction comes into play . in general , our understanding of the selection of the de wave vector is not fully developed . however , we can exclude , that the laser spot sizes of pump ( @xmath72 ) or probe ( @xmath73 ) enforce a selection , since the observed de wave lengths are of the order of only very few microns and we do not observe any dependencies on the pump or probe beam diameter . as described , the de mode dominates the spectra up to critical fields as high as @xmath74 . around that value , an explicit deviation from the theoretical dispersion arises ( blue line and points in the bottom row of ) . a distinct modification of the de mode intensity takes place which for sufficiently high fields ultimately results in a peculiar characteristic evident in . namely the de mode merges into the kittel mode . this leads to a very strong increase of the total fourier power , i.e. precession amplitude , as seen in the non - normalized data ( upper row of ) . the propagating surface mode seems to lock to the frequency of the kittel and pssw modes for magnetic fields higher than @xmath75 . above that value , the peaks can be attributed to the dispersion of the uniform precession by eq . [ eq : kittel ] . the fits ( white lines ) yield values for the effective anisotropy @xmath76 around @xmath77 showing no monotonous trend with the thickness . to discuss this unusual observation we want to remind the reader , that the highly asymmetric initial excitation profile unfavors the uniform precession as a magnetic eigen mode . one possibility to understand its observation is a locking of the de mode to the @xmath78 homogenious precession originating from an energy transfer between the two dynamic modes . only if the propagating surface waves would couple to the localized kittel mode the amplitude of the magnetic precession could be that drastically increased . a reasonable starting point to discuss this idea is to draw parallelities to a model by slavin and coworkers . they proved , that phase locking of spin - torque nano - oscillators can take place.@xcite in their publication , the authors use a non - linear set of equations of motion for two coupled oscillators and can explain the experimental results presented in refs . once the free running frequency mismatch between the two oscillators is smaller than a threshold value @xmath79 ( eq . ( 9 ) in ref . [ ] ) , they phase lock to a mutual frequency . the mechanism is twofold : coupling can be either mediated by spin waves radiated into a common magnetic layer or a dipole field created by the oscillators . either way , one of the crucial parameters is the spatial separation @xmath80 of the two contacts , tuning the overlap of spin - wave power in real space ( fig . 1 in ref . [ ] ) . transferring these findings to our observation , we could speculate that an increased overlap of the magnetic modes in fourier space leads to a sudden interaction such that locking takes place . figure [ fig:03-fft ] illustrates this hypothesis : two fourier spectra from measurements recorded on nickel with a thickness of @xmath81 are shown . at @xmath59 clearly two precessional modes can be distinguished , whereas at @xmath82 only one smooth peak with nearly identical width is found . the two modes observed , namely uniform precession and de surface waves , are separated in fourier space by several ghz at low external fields . however , if @xmath50 is increased in the experiment , the separation reduces and at one point approaches the value of the linewidth , which implies an overlap in the frequency domain ( hatched area in ) . above a threshold overlap the frequencies seem to lock and the two modes can not be distinguished anymore . moreover , a single smooth peak in the fourier spectrum is observed . concerning the evolution with increasing magnetic field , a step - like change of the de frequency is apparent in the fourier spectra in and hence , the term mode locking seems justified . equivalent to the damping time constant @xmath83 in the time domain is the linewidth in the fourier domain . from @xmath83 apparent in the time - resolved moke spectra one expects the width of the peak after fourier transformation to be @xmath84 at 50% of maximum power ( fwhm , full width half maximum ) . for nickel the expected linewidth therefore is @xmath85 , using @xmath86 , which has been extracted from the time - resolved spectra . due to the strong damping in nickel ( implying a rather large line width ) considerable overlap can occur . the interplay with @xmath87 tuning the de and kittel frequency mismatch leads to possible phase locking ( see ) . of the kittel mode . the external magnetic field can be used tune the observed characteristics between dipolar surface waves and uniform precession . ( b ) determination of the critical magnetic field @xmath75 for different fluences @xmath88 and @xmath89 , respectively . @xmath90 is plotted after normalization to the maximum of the respective fourier spectrum for a given external field , @xmath91 . the result can be regarded as a measure of mode population and a linear increase is found prior to saturation . the critical field can be extracted as depicted by the vertical green lines . in the inset , corresponding results for all fluences are included , giving a clear trend towards smaller @xmath75 for increasing @xmath92 . additionally , the kittel amplitude as induced by different pump fluences is shown for @xmath93 ( gray points ) . ] a test for this hypothesis based on non - linearity is the influence of the ( optical ) pumping power that excites the magnetization dynamics . corresponding experiments were performed on a separate continuous nickel film which was @xmath94 thick and was prepared as described above . for the measurements , the pumping fluence was varied over one decade ranging from @xmath95 to @xmath96 . the resulting fourier power spectra are presented in and reveal that the nature of the magnetic modes excited by the pump pulses does not significantly change . to give a better insight into the fluence dependence , all power spectra in the top row are identically color - scaled . this shows an increase of the spin - wave amplitude with the fluence ( from left to right ) as expected when considering the heat - induced disorder triggering the oscillation.@xcite this increase is quantitatively given in the inset of ( b ) , where the kittel amplitude at @xmath93 has been plotted versus the pumping fluence ( gray points ) . in this respect , the absolute value of the amplitude is of great interest . it can be derived from the dynamic magnetization by calibration of the experiment for a given saturation magnetization : a comparison of the magneto - optical signal at zero time delay with and without pump - induced demagnetization at @xmath97 and @xmath93 results in a precessional amplitude of @xmath98 , corresponding to an angle of @xmath3 . a feature of that requires a closer look is the somewhat changing critical field for the transition from surface to uniform characteristics . figure [ fig:05-population](b ) details the corresponding analysis and gives the respective results in the inset . starting from the fit of the kittel dispersion to the data ( white lines in the bottom row of ) the amplitude of the kittel mode for each magnetic field is extracted from the data . in the normalized case , this amplitude equals @xmath99 if @xmath100 and below the critical field a monotonous , in good approximation linear behavior is found . by a linear fit , the point at which the kittel amplitude reaches @xmath99 , i.e. above which the kittel mode is the dominant mode of precession can be determined . we define this field value as the critical field @xmath101 ( also marked by vertical lines in ( b ) ) . the above described procedure is explicitly given in ( b ) for the fluences @xmath95 and @xmath89 , respectively . in the inset the overall decrease of @xmath75 for increasing @xmath92 is shown . together with the threshold - like onset of the uniform precession in ( a ) and the accompanying rapid increase of precession amplitude our initial hypothesis of a non - linear transfer of energy seems to be verified . in conclusion , we have identified the excitation of the uniform precession ( kittel mode ) as well as magnetostatic surface waves ( de mode ) in asymmetrically pumped nickel films . below a critical magnetic field @xmath101 the dipolar de modes dominate the precession , whereas a step - like feature in the dispersion @xmath102 indicates the onset of the population of the @xmath78 mode above @xmath101 ( ) . the observation of the de surface waves for @xmath103 can be well understood by considering the optical excitation mechanism . a concept was developed taking into account the amplitude profile of the de modes as well as the asymmetric excitation profile stemming from the laser pump pulses . however , the understanding of the selection of relaxation channels in terms of the selection of the de wave vector is a challenge still to be met . the transistion to uniform precession for @xmath100 was discussed in terms of a non - linear mechanism transferring energy between the two eigen modes of the ferromagnet . only a non - linear process allows to explain the dependence of the transition field @xmath101 on the pumping power . in this context the high energy of the pump pulses driving the system far out of equilibrium leading to large precession angles should be mentioned . in total , the population of either the kittel or the de mode can be tuned by both the external field and the pumping fluence . understanding the underlying mechanism might prove crucial on the way to spin - logic devices , especially with the interconversion of electrical , optical or magnetic signals in mind . one of the authors ( j.h . ) acknowledges financial support through grant academy of the academy of sciences of the czech republic ( kan400100653 ) . also , fruitful discussions with a. slavin and b. hillebrands are gratefully acknowledged . , , , , , , * * , ( ) , issn , http://search.ebscohost.com / login.aspx?direct = true&db = pbh&an=% 18301425&site = ehost - live[http://search.ebscohost.com / login.aspx?direct = true&db = pbh&an=% 18301425&site = ehost - live ] . , , , , * * , ( ) , http://www.sciencedirect.com / science / article / b6tjj-46fx65x-4t% /2/a4275019339a05e20b314317a82cd8ae[http://www.sciencedirect.com / science / article / b6tjj-46fx65x-4t% /2/a4275019339a05e20b314317a82cd8ae ] .	the spin - wave relaxation mechanisms after intense laser excitation in ferromagnetic nickel films are investigated with all - optical pump - probe experiments . uniform precession ( kittel mode ) , damon - eshbach surface modes and perpendicular standing spin waves can be identified by their dispersion @xmath0 . however , different to other ferromagnets @xmath0 deviates from the expected behavior . namely , a mode discontinuity is observed , that can be attributed to a non - linear process . above a critical field the power spectrum reveals a redistribution of the energy within the spin - wave spectrum populated .
progress in many areas of science and its application is fueled by the ongoing progress to measure and structure small objects . in many cases , light is used as a primary tool for reading or writing . but since light is subject to diffraction , a straightforward implementation is restricted to structures of order of the involved wavelength @xcite . different methods have been invented to surpass this limit , such as near - field imaging @xcite , techniques based on the selective addressing of nearby particles @xcite , resolution enhancement due to non - classical effects @xcite , multiphoton spectroscopy @xcite , quantum lithography with classical fields @xcite , or position - dependent dark states @xcite . among the most fundamental problems in this area is the measurement of the distance between two nearby quantum particles such as atoms . it has been recognized that a precise determination of the interparticle distance is possible down to distances far below the wavelength of the employed light based on their mutual interaction . for small distances , the atoms are coupled by the dipole - dipole interaction , which modifies the optical properties of the system @xcite . this was confirmed in a recent experiment @xcite , and it was found that the resonance fluorescence exhibits characteristic features which enable one to determine the relative distance over a large range of small distances @xcite . the resonance fluorescence has the advantage that it can be observed in the far field , and distance determination via fluorescence is not affected by the usual resolution limitations since the distance information is encoded in the frequency spectrum of the emitted light . similar ideas have also been used for the localization of single particles @xcite . the existing distance measurement techniques based on the dipole - dipole interaction , however , are restricted to two two - level atoms in specific geometries , such as aligned along the propagation direction of the exciting laser field . in most practical cases , however , the relative orientation of the two nearby atoms is unknown , and for similar reasons , it is equally difficult to measure the relative orientation as the distance . thus the question arises , whether the ideas of @xcite can be extended to the case of arbitrary orientation . it turns out that it is not meaningful to study the system of two two - level atoms in the case of arbitrary orientations . the reason for this is the appearance of dipole - dipole couplings between orthogonal transition dipole moments ( ddotdm ) in more general geometries @xcite . the electric field emitted by one of the particles has not only a component corresponding to the emitting transition dipole moment , but also a component along the interparticle distance vector . the projection of the latter field component on a transition dipole moment in the second atom can be non - zero even if it is orthogonal to the emitting dipole @xcite . in a real atom with magnetic level structure , these ddotdm lead to the population of excited states even if they are not driven by the external laser field . thus , the two - level approximation breaks down , and correct predictions can only be expected if the theoretical modelling includes complete zeeman manifolds including all occurring dipole dipole couplings @xcite . motivated by this , here we study the determination of relative distance and orientation of two nearby atoms in arbitrary geometry . the atoms are driven by a single resonant standing - wave laser field , and we make use of the far field resonance fluorescence intensity and spectrum as observables . each atom is modelled as a four - level system with one ground state ( total angular momentum zero ) and three excited states ( total angular momentum one ) , including all relevant dipole - dipole couplings occurring in arbitrary geometries . we start by analyzing the distance determination for the case of a known orientation , and present dressed - state interpretations of the obtained resonance fluorescence spectra in various cases of relevance . next , we describe a method to determine the interparticle distance for arbitrary orientation , which works as long as the particles are sufficiently close to each other . finally , we discuss methods to determine the relative orientation of the two particles , focusing on the two cases of atoms confined in a planar waveguide and atoms on a surface . ( color online ) scheme for the determination of relative distance and orientation of two nearby atoms . the atoms @xmath0 are driven on one transition by a standing wave laser field with rabi frequency @xmath1 and scatter light , which is registered in the far field with a detector . the interatomic distance vector @xmath2 is arbitrary , as shown in the left inset . the right inset shows the level structure of each atom . each atom has a single ground state ( zero angular momentum ) and three excited states ( angular momentum 1 ) . ] our system consists of two identical nearby atoms located at @xmath3 ( @xmath4 ) as shown in fig . [ system ] . each atom is modelled as a four - level system with one ground and three excited states , modelling the complete magnetic substructure of a transition from an @xmath5 state ( angular momentum 0 ) to a @xmath6 state ( angular momentum 1 ) . the hamiltonian of the two atoms coupled to the surrounding vacuum field is given by @xmath7 where @xmath8 in these equations , @xmath9 is the hamiltonian that describes the free evolution of the bi - atomic system . we set the energy of the ground state to zero , and the energies of the excited atomic states @xmath10 ( @xmath11 ) are @xmath12 . the raising and lowering operators on the @xmath13 @xmath14 transition of atom @xmath15 @xmath16 are @xmath17 the hamiltonian of the vacuum field is described by @xmath18 , with @xmath19 and @xmath20 as the field annihilation and creation operators . @xmath21 describes the interaction of the atom with the vacuum field in the dipole approximation with the vacuum field @xmath22 given by @xmath23 @xmath24 is the wave vector , @xmath25 the polarization , @xmath26 the frequency of a field mode , and @xmath27 the quantization volume . we use the wigner - eckart theorem @xcite to determine the electric dipole moment operator of atom @xmath15 which is given by @xmath28 and the dipole moments @xmath29 are given by the matrix elements of the electric dipole moment operator @xmath30 as @xmath31 @xmath32 , and @xmath33 denotes the reduced dipole matrix element . note that the dipole moments @xmath34 are independent of the atomic index @xmath15 since the two atoms are identical . the vector that defines the relative position of the atoms in spherical coordinates is given by @xmath35 using standard methods , the system master equation evaluates to @xcite @xmath36-\dfrac{\iota}{\hbar}[h_{\omega},\varrho]+\mathcal{l}_{\gamma}\varrho.\ ] ] the hamiltonian @xmath37 arises from the coherent part of the dipole - dipole interaction and is given by @xmath38 the incoherent part of eq . ( [ master ] ) is given by @xmath39 in the above equations , coefficients @xmath40 and @xmath41 with @xmath42 represent dipole - dipole couplings between a transition dipole of one atom and the corresponding parallel dipole of the other atom , whereas terms with @xmath43 correspond to dipole - dipole couplings between orthogonal transition dipole moments in the two atoms . the coefficients @xmath40 and @xmath41 can be calculated as @xmath44\ , , \label{oddes}\\ % \gamma_{ij}=\dfrac{1}{\hbar}[\textbf{\textit{d}}_{i}^{t } \textit{\rm{im}}(\overleftrightarrow{\mathcal{x}})\textbf{\textit{d}}_{j}^{\ast}]\,.\label{oddgam}\end{aligned}\ ] ] here , @xmath45 [ @xmath46 represents the real [ imaginary ] part of the tensor @xmath47 whose components @xmath48 ( @xmath49 ) are given by @xmath50e^{\iota\eta}.\end{aligned}\ ] ] @xmath51 is the kronecker delta symbol and @xmath52 , with the approximation @xmath53 , where @xmath54 is the mean transition frequency . from eq . ( [ oddes ] ) , the different dipole - dipole coupling constants @xmath40 evaluate to @xmath55\sin^{2}\theta e^{-2\iota\phi}\,,\\ % \omega_{11}&=3\dfrac{\gamma}{8\eta^{3}}\{[3\eta^{2}-1+(\eta^{2}-3)\cos2\theta]\cos\eta\nonumber\\ & -\eta(1 + 3\cos2\theta)\sin\eta\}\ , , \label{om11}\\ % \omega_{21}&=-\sqrt{2}\cot\theta\omega_{31}e^{\iota\phi}\,,\\ % \omega_{22}&=\omega_{11}-(2 \cot^{2}\theta-1)\omega_{31}e^{2\iota\phi}\ , , \label{om22}\\ % \omega_{32}&=-\omega_{21 } , \quad \omega_{33}=\omega_{11}\,.\end{aligned}\ ] ] the corresponding incoherent coupling constants @xmath41 follow from eq . ( [ oddgam ] ) as @xmath56\sin^{2}\theta e^{-2\iota\phi}\,,\\ % \gamma_{11}&=3\dfrac{\gamma}{8\eta^{3}}\{[3\eta^{2}-1+(\eta^{2}-3)\cos2\theta]\sin\eta\nonumber\\ & + \eta(1 + 3\cos2\theta)\cos\eta\}\,,\\ % \gamma_{21}&=-\sqrt{2}\cot\theta\gamma_{31}e^{\iota\phi}\,,\\ % \gamma_{22}&=\gamma_{11}-(2 \cot^{2}\theta-1)\gamma_{31}e^{2\iota\phi}\,,\\ % \gamma_{32}&=-\gamma_{21 } , \quad \gamma_{33}=\gamma_{11}\,.\end{aligned}\ ] ] some examples illustrating the dependence of the coupling constants @xmath40 on @xmath57 or @xmath58 are shown in fig . [ fig - coupling ] . the total spontaneous decay rate of each individual atom is given by @xmath59 , where @xmath60 and we have again used the approximation @xmath61 . ( color online ) ( a ) magnitude of few orthogonal dipole - dipole coupling constants . the parameters are @xmath62 and @xmath63 . the red solid curve shows @xmath64 , while the blue dashed curve depicts @xmath65 . ( b ) magnitude of few parallel dipole - dipole coupling constants at @xmath66 . the red solid line shows @xmath67 , and the dashed blue line depicts @xmath68.,title="fig : " ] ( color online ) ( a ) magnitude of few orthogonal dipole - dipole coupling constants . the parameters are @xmath62 and @xmath63 . the red solid curve shows @xmath64 , while the blue dashed curve depicts @xmath65 . ( b ) magnitude of few parallel dipole - dipole coupling constants at @xmath66 . the red solid line shows @xmath67 , and the dashed blue line depicts @xmath68.,title="fig : " ] we finally include an external driving laser field , which we assume to be polarized along the z - axis . its electric field is given by @xmath69 where @xmath70 denotes the amplitude , @xmath71 the frequency , and @xmath72 is the polarization of the field , and @xmath73 denotes the complex conjugate . the wave vector @xmath74 with wave number @xmath75 points along the positive x - axis . in a suitable interaction picture , we obtain @xmath76-\dfrac{\iota}{\hbar}[h_{\omega},\tilde{\varrho}]+\mathcal{l}_{\gamma}\tilde{\varrho}\,,\ ] ] with @xmath77 here , the detunings @xmath78 . the interaction of the system with the external laser field in the electric dipole and the rotating wave approximation is described by @xmath79}\,.\ ] ] the position dependent rabi frequencies are given by @xmath80 where @xmath81 ( @xmath0 ) . due to its polarization , the laser field couples only to the @xmath82 transitions in the two atoms . our observables are the resonance fluorescence intensity and the resonance fluorescence spectrum of the light emitted by the two atoms . the spectrum of resonance fluorescence up to a geometrical factor is determined by real part of the fourier transform of the two time correlation function of the electric field @xcite , @xmath83 in this equation , @xmath84 @xmath85 $ ] denotes the positive [ negative ] frequency part of the electric field operator and the subscript @xmath86 refers to the steady state . at a point @xmath87 in the far - field zone , the negative frequency part of the electric field operator evaluates to @xcite @xmath88 where @xmath89 is the retarded time and @xmath90 . the first term denoting the free field can be neglected if the point of observation lies outside the driving field . the resonance fluorescence intensity is given by the one - time normally ordered correlation function of the electric field operators , @xmath91 in the following , the point of observation is assumed to be along the @xmath92 or @xmath93 direction for the resonance fluorescence spectrum , and along the @xmath94 direction for the resonance fluorescence intensity . evaluating the cross products in eq . ( [ negfield ] ) , we find that our choice of observation direction enables us to separate linearly polarized light emitted on transitions @xmath95 from the circularly polarized light emitted on transitions @xmath96 ( @xmath97 ) by means of a polarization analyzer . we designate linearly [ circularly ] polarized spectra as @xmath98 [ @xmath99 ones , respectively . the resonance fluorescence spectrum eq . ( [ spectrum ] ) emitted by the two atoms in general is rather complicated , but it simplifies considerably in certain parameter cases , as it was found already in the case of two nearby two - level systems @xcite . in the following , we will in particular refer to the case of either small or large interatomic spacing , on a length scale given by the involved transition wavelength . for small distance , the coherent part of the dipole - dipole interaction dominates the system dynamics , with corrections due to the much weaker laser field rabi frequencies . in the opposite case of larger separation , the rabi frequencies dominate , with corrections from the dipole - dipole interaction . spectra for situations in which the dipole - dipole interaction and the rabi frequencies are comparable usually can not be interpreted in a straightforward way . in these cases , the driving field intensity can be increased or decreased in order to evolve in one of the two simpler cases . in the following , we will make use of this general observation , and present our results in two steps . first , we will describe methods to determine the interatomic separation in various cases of interest . second , we will discuss the determination of the relative orientation of the two atoms . for the numerical analysis , we assume that @xmath100 . our measurement techniques , however , also apply to other values of @xmath101 . a special case arises if one of the atoms is at a node of the standing wave field . such situations can be circumvented by shifting the phase of the standing wave slightly . we also assume the resonance condition , _ i.e. _ , @xmath102 , see eq . ( [ ham - laser ] ) . ( color online ) ( a ) incoherent spectrum of resonance fluorescence for the case of larger atomic distances . the parameters are @xmath103 , @xmath104 , @xmath105 and @xmath106 . the side peaks are located at the rabi frequencies , @xmath107 and @xmath108 . ( b ) the dressed state representation of the system in ( a ) showing the ac - stark splitting of the two atoms . the splitting in each atom corresponds to the respective position dependent rabi frequency . , title="fig : " ] ( color online ) ( a ) incoherent spectrum of resonance fluorescence for the case of larger atomic distances . the parameters are @xmath103 , @xmath104 , @xmath105 and @xmath106 . the side peaks are located at the rabi frequencies , @xmath107 and @xmath108 . ( b ) the dressed state representation of the system in ( a ) showing the ac - stark splitting of the two atoms . the splitting in each atom corresponds to the respective position dependent rabi frequency . , title="fig : " ] as a first step , we will present results for the situation of a fixed , known orientation . in particular , for simplicity , we analyze the case @xmath109 . then , the orthogonal dipole - dipole couplings @xmath110 and @xmath111 ( @xmath97 ) vanish . thus , the population is trapped only in the levels @xmath112 and @xmath113 , and the system essentially reduces to that of two two - level atoms @xcite . this simple case , however , suffices for the main aim of this section , which is to extend previous results by interpreting each case in detail in terms of the corresponding dressed state picture . this insight will enable us to explain our later main results for general geometries . in principle , the results in this section also generalize to more complicated known orientations . while then the dressed - state analysis is in complete analogy to our discussions here , the analytical expressions are considerably more complicated . in any case , if the orientation is known , a numerical fit of the measured spectrum leads to the desired distance information . in the following section [ unknown ] , we will extend our analysis to arbitrary orientations . ( color online ) ( a ) incoherent spectrum of resonance fluorescence for intermediate distances . the parameters are @xmath114 , @xmath105 , @xmath104 and @xmath115 , such that the laser rabi frequencies are larger than the relevant dipole - dipole coupling constants . the centre of the side band structures are located at the respective position dependent rabi frequencies , @xmath116 and @xmath117 . ( b ) dressed - state representation . ( i ) shows the system without couplings . ( ii ) includes the dominant laser field couplings . ( iii ) in addition includes corrections due to the dipole - dipole interaction . , title="fig : " ] ( color online ) ( a ) incoherent spectrum of resonance fluorescence for intermediate distances . the parameters are @xmath114 , @xmath105 , @xmath104 and @xmath115 , such that the laser rabi frequencies are larger than the relevant dipole - dipole coupling constants . the centre of the side band structures are located at the respective position dependent rabi frequencies , @xmath116 and @xmath117 . ( b ) dressed - state representation . ( i ) shows the system without couplings . ( ii ) includes the dominant laser field couplings . ( iii ) in addition includes corrections due to the dipole - dipole interaction . , title="fig : " ] for larger interatomic separation ( about @xmath118 ) , the dipole dipole interaction is almost negligible and the two atoms can be considered independent . a typical spectrum is shown in fig . [ fig - large](a ) , and consists of five peaks . the middle peak occurs at @xmath119 , and the two side peak doublets are symmetrically located at @xmath120 . an interpretation of the peak structure in terms of the system dressed states is shown in fig . [ fig - large](b ) . it turns out that the sidebands are shifted by the position - dependent rabi frequencies experienced by the two atoms , respectively . thus , the peak positions give a direct measure of the position of each atom and hence the interatomic separation can be calculated using the expressions for the position - dependent rabi frequencies eq . ( [ rabifreq ] ) @xcite . the atomic levels in atom @xmath15 split by the amount @xmath121 , just as in the mollow spectrum @xcite . note that the two split doublets may coincide , for example , for @xmath122 , since then @xmath123 . for intermediate interatomic distances ( approx . @xmath124 ) , the dipole dipole interaction becomes relevant . if the parameters are such that the driving field rabi frequency is comparable to the dipole - dipole interaction , then the spectrum obtained is difficult to interpret . this , for example , occurs if the distance is decreased to @xmath114 in fig . [ fig - large](a ) . we thus increase the rabi frequency of the driving laser field and obtain the spectrum shown in fig . [ fig - middle](a ) . it can be observed that the sidebands are split by the dipole - dipole interaction . the corresponding dressed - state picture is shown in fig . [ fig - middle](b ) . this picture is obtained by evaluating the eigenstates and eigenvalues of the interaction picture hamiltonian in the limit @xmath125 . we consider the case @xmath126 and assume @xmath127 . in part ( i ) of fig . [ fig - middle](b ) , there is no interaction between the two atoms , such that four atom - field states are degenerate . in ( ii ) , we include the interaction with the strong driving field , which splits the level scheme into four dressed states . the corresponding dressed states are listed in tab . [ my ] . finally , in part ( iii ) , the weak dipole - dipole interaction is added , which further shifts the eigenenergies of all dressed states by approximately @xmath128 . ( color online ) incoherent spectrum of resonance fluorescence in the small distance case , with laser field rabi frequencies negligible compared to the dipole - dipole coupling constants . the parameters are @xmath129 , @xmath105 , @xmath104 and @xmath130 . the side peaks occur at @xmath131 . ( b ) dressed state representation . the energies of the symmetric @xmath132 and the antisymmetric @xmath133 states are shifted by the dipole dipole coupling parameter @xmath134 . , title="fig : " ] ( color online ) incoherent spectrum of resonance fluorescence in the small distance case , with laser field rabi frequencies negligible compared to the dipole - dipole coupling constants . the parameters are @xmath129 , @xmath105 , @xmath104 and @xmath130 . the side peaks occur at @xmath131 . ( b ) dressed state representation . the energies of the symmetric @xmath132 and the antisymmetric @xmath133 states are shifted by the dipole dipole coupling parameter @xmath134 . , title="fig : " ] .[my]eigenvectors and eigenvalues of the interaction hamiltonian of two two - level atoms in the limit @xmath135 . [ cols="<,<,<",options="header " , ] the methods presented so far allow for a determination of the interatomic distance if the orientation of the two atoms is known and fixed . often , however , the orientation is unknown . therefore , in this section , we turn to our main results , and present a method to obtain the interatomic distance for arbitrary orientations . this method can be applied if the interatomic distance is sufficiently small , such that the dipole - dipole interaction dominates the system dynamics . fortunately , this usually is exactly the parameter range in which a distance determination is desired . in order to explain the method , we first imagine the two atoms without any driving fields . then , it turns out that the eigenenergies of the dressed states are independent of the orientation of the two atoms @xcite . an interpretation of this fact is that without external field , there is no preferred direction in space , such that the energies can not depend on the orientation . since the spontaneously emitted light is emitted at frequencies corresponding to the dressed state energies , it follows that the positions of the peaks in the fluorescence spectrum of the atoms are unaffected by the orientation of the atoms . this property is approximately preserved if the atoms are driven by a weak driving field , which has a rabi frequency much smaller than the dipole - dipole couplings . we therefore find that at small distance and weak driving , the resonance fluorescence spectrum has peak positions independent of the alignment of the two atoms . only the relative widths and heights of the spectral features change with the orientation . it is important to note that these properties of the two - atom system are only described correctly if all dipole - dipole couplings are included in the modelling @xcite . this is the reason why we included complete zeeman manifolds in our analysis . ( color online ) incoherent spectrum of resonance fluorescence for small distance and weak driving fields . the parameters are @xmath129 , @xmath136 , @xmath137 and @xmath130 . ] it remains to deduce the interatomic distance from the peak positions in the resonance fluorescence spectrum . for this , we again analyze the eigenvalues of the interaction hamiltonian , which determine the peak positions . since the peak positions and thus these eigenvalues are independent of the orientation , it suffices to evaluate analytic expressions for the peak positions in a simple configuration . investigating the eigenvalues for @xmath138 , and assuming @xmath139 , we find that the eigenenergies are given by @xmath140 , @xmath141 , and @xmath142 . an example for this is shown in fig . [ arb - orient ] . the four side peaks are located at @xmath143 and @xmath144 , and the interatomic distance can be gained from both coupling constants via eq . ( [ om11 ] ) and ( [ om22 ] ) . as can be seen from fig . [ fig - coupling ] , for @xmath122 , the coupling constant @xmath145 is larger than @xmath146 for small interatomic distances @xmath58 . thus , the inner [ outer ] peaks in fig . [ arb - orient ] correspond to @xmath146 [ @xmath145 ] . due to the dependence of the amplitude of the spectral peaks on the orientation , the peaks at @xmath143 visible in fig . [ arb - orient ] may be suppressed . for example , in fig . [ fig - small](a ) , we found only a single pair of sidebands corresponding to @xmath144 . this can be understood by observing that these peaks at @xmath144 correspond to states @xmath148 populated by the driving laser field , while states @xmath149 and @xmath150 are only populated in certain geometries . we end this section by noting that in the case of large interatomic separation , the dipole - dipole interaction vanishes , such that the interatomic coupling also becomes independent of the orientation . the driving field rabi frequencies experienced by the two atoms , however , may not be the same , as they depend on the scalar product @xmath151 . thus , by applying a driving field with larger rabi frequency , it is possible to measure the position of the atoms projected on the propagation axis of the driving field . for arbitrary orientations , however , a single measurement of this type does not allow to deduce the interatomic distance since the position transverse to the wave vector remains unknown . we thus conclude that by applying weak driving fields , the interatomic distance can be measured from a pair of nearby atoms independent of their mutual orientation , as long as the dipole - dipole interaction is strong enough to dominate the system dynamics . ( color online ) setup with the two nearby atoms confined inside a two - dimensional waveguide . the atoms are located in the @xmath152-@xmath92 plane . ] so far , we have discussed techniques for the measurement of the interatomic distance , and have demonstrated how the distance can be measured independent of the orientation of the interparticle distance vector . in this section , we augment our analysis by discussing the determination of the relative orientation of the two atoms . we discuss two different cases , corresponding to two different methods to determine the orientation . first , we discuss the case of unknown @xmath153 , assuming @xmath105 . this case corresponds to an effective two - dimensional geometry of the system which can be realized , e.g. , by embedding the atoms in a planar matter waveguide @xcite . in this case , the orientation is deduced from the @xmath153-dependent peak positions in the fluorescence spectrum induced by the driving laser field . second , we study the case of unknown @xmath57 and @xmath63 . this corresponds to atoms on a surface , driven by a laser field propagating perpendicular to the surface . in this case , we will determine @xmath57 via the resonance fluorescence intensity emitted in a particular direction . ( color online ) incoherent spectrum of resonance fluorescence for two atoms in a geometry as shown in fig . [ fig-2d ] . the parameters are @xmath154 , @xmath105 , @xmath155 , and @xmath156 . ] in this section , we assume that the two atoms are confined in the @xmath152-@xmath92 plane ( @xmath105 ) as shown in fig . [ fig-2d ] , as it is the case , for example , in a planar waveguide . since such a setup may also constrain the observation direction , we assume detection in a direction anti - parallel to the incident driving field , i.e. , along the @xmath93 direction . this way , the spectrum can be measured without background from the incident laser field . in this geometry , the coupling constants @xmath157 and @xmath158 are zero for all values of @xmath153 . since only the second transition @xmath82 is driven , the populations of the levels @xmath149 and @xmath150 are zero for the whole range of @xmath153 . the parallel coupling constants @xmath159 and @xmath134 are independent of @xmath153 . nevertheless , the spectra depend strongly on @xmath153 because of the @xmath153-dependence of the rabi frequency @xmath160 . the obtained spectra are identical if @xmath153 is replaced by @xmath161 . from the results of sec . [ unknown ] it is clear that the peak positions in the resonance fluorescence spectrum can not be used to determine the orientation as long as the dipole - dipole interaction dominates the dynamics . therefore , we apply stronger driving fields , such that the external driving dominates the dynamics . in this case , for most values of @xmath153 , a typical spectrum obtained is shown in fig . [ 2d - strong ] . using the results of sec . [ fixed ] , we conclude that the two doublets on each side corresponding to mollow sidebands at the two rabi frequencies experienced by the atoms , split by the dipole - dipole interaction . the doublets can thus be used to approximately read off the two position - dependent rabi frequencies . assuming that the distance is known from a measurement with a weaker driving field as described in sec . [ unknown ] , the components of the position vectors of the individual atoms as well as the relative alignment of the atoms along the laser can be found out by using the position dependent rabi frequencies @xmath162 as follows : @xmath163\right\}\,.\ ] ] in fig . [ 2d - strong ] , the peaks in the spectrum of resonance fluorescence occur at approximately @xmath140 , @xmath164 , @xmath165 , @xmath166 , and @xmath167 . considering the mean values of the peak separation in the inner side band doublets as @xmath168 and that in the exterior side band doublets as @xmath160 , from eq . ( [ eq - phi ] ) we obtain @xmath169 as the value of @xmath153 , which deviates from the true value by about @xmath170 . this deviation can be attributed to the imperfect determination of the rabi frequencies as the mean value of the two peaks in the doublets . increasing @xmath153 from the value that has been used in fig . [ 2d - strong ] , the position - dependent rabi frequencies change , until the sideband doublets corresponding to the two rabi frequencies start to overlap , as the two position - dependent rabi frequencies approach each other . in this case , it is difficult to estimate @xmath153 directly from the spectrum , since a clear identification of the different peaks is not obvious . one strategy is to increase the driving field intensity . since the peak separation @xmath171 is proportional to @xmath172 , this increase eventually leads to a splitting larger than the line widths of the involved peaks , such that an identification becomes possible . in any case , it can be concluded from overlapping peaks that @xmath153 is close to @xmath173 or @xmath174 , since then @xmath175 . in summary , the relative orientation of the two atoms with respect to the laser can be determined using stronger laser fields . this works well if the position dependent rabi frequencies are different from each other , since then the spectral lines are well separated . this is the case for @xmath153 not close to @xmath173 or @xmath174 . accordingly , if the corresponding spectral peaks overlap , it can be concluded that @xmath153 is close to @xmath173 or @xmath174 . ( color online ) fluorescence intensity emitted on the @xmath176-transitions observed in @xmath94 direction from two atoms on a surface . @xmath177 is the relative angle between laser polarization and interatomic distance vector . the green dashed lines indicate the four possible values of @xmath177 corresponding to a possible measured intensity indicated by the solid red line . the parameters are @xmath178 , @xmath63 and @xmath115 . ] in this section , we consider the case of two atoms on a surface , driven by a laser field propagating perpendicular to the surface . thus , @xmath153 is fixed to @xmath173 , while @xmath57 is unknown . in this case , the two rabi frequencies experienced by the atoms are equal and independent of @xmath58 and @xmath57 . therefore , in contrast to the previous sec . [ unknown - phi ] , here , we determine the value of @xmath57 with the help of the resonance fluorescence intensity . in particular , we consider the @xmath176-intensity emitted by the dipoles @xmath179 and @xmath180 , measured by a detector placed in @xmath94 direction since there is no @xmath176-spectrum in @xmath92 direction . it turns out that the configuration is symmetric in the sense that a rotation of the laser polarization and the detectors around the @xmath152 direction is equivalent to a corresponding rotation of the interatomic distance vector . therefore , the measured resonance fluorescence intensity depends only on the relative angle @xmath177 between the laser polarization direction and the orientation of the two atoms @xmath57 on the surface . in fig . [ int - theta ] , we show this resonance fluorescence intensity versus the relative angle @xmath177 . a plot like this can either be recorded by rotating the sample in the @xmath92-@xmath94 plane , or by rotating the polarization vector of the laser field around its propagation axis . from fig . [ int - theta ] , we find that the intensity is symmetric around @xmath181 , and it is easy to see that one value of @xmath176 intensity corresponds to at most four values of @xmath177 . these four values of @xmath182 can be roughly divided into the four ranges @xmath183 , @xmath184 , @xmath185 and @xmath186 , respectively . this can be understood by noting that the orthogonal coupling constants responsible for the @xmath176 intensity are the same for orientations @xmath57 , @xmath187 , @xmath188 and @xmath189 . at @xmath190 , the @xmath176 intensity is zero since the orthogonal dipole dipole coupling constants @xmath158 and @xmath157 vanish at these points , see fig . [ fig - coupling](a ) . thus there is no population in states @xmath149 and @xmath150 , and the intensity of light emitted from these states is zero . the points of zero @xmath176 intensity @xmath191 correspond to situations in which the polarization vector is parallel , perpendicular , or anti - parallel to @xmath2 , respectively . since these values of @xmath177 with vanishing intensity can easily be identified , they allow to determine @xmath57 from the amount of sample or driving field polarization rotation required to reach these values . in particular , the symmetry point @xmath181 is well - suited for such a measurement . we have discussed methods to measure the relative distance and orientation of two nearby atoms in arbitrary geometry . our methods are based on the driving of the two atoms with a standing wave field , and on detection of the resonance fluorescence intensity and spectrum in the far field . the distance and orientation information is encoded in the scattered light via the position - dependent rabi frequencies and via the distance- and orientation - dependent dipole - dipole couplings . since unlike in previous studies , we consider the case of arbitrary orientation , the atoms must be described using complete zeeman manifolds in order to correctly model all relevant dipole - dipole couplings between parallel dipole moments as well as between orthogonal ones . as preliminary work , we have analyzed the fluorescence spectra in particular known geometries , in order to identify dressed - state interpretations in the various limiting cases of interest . these in particular are the case of dominating laser - induced dynamics perturbed by the dipole - dipole interaction , and the case of dominating dipole - dipole coupling modified by the presence of a weaker laser field . next , we have shown that the case of dominating dipole - dipole interaction enables one to measure the distance between two nearby particles independent of the relative orientation . the reason for this is that the eigenvalues of the total hamiltonian describing the dynamics , and thus the position of the system dressed states , are independent of the orientation if the two atoms are undriven . we found that a weak driving field allows to probe these dressed states without perturbing the independence on the orientation . finally , we discussed the measurement of the relative orientation . we presented two methods . the first is based on the position - dependent rabi frequencies , which under certain conditions reveal the orientation of the two particles . the second method is based on the measurement of the resonance fluorescence intensity in a particular direction . this intensity is a measure for the population in the excited states not driven by the laser field , and therefore a signature for the magnitude of the dipole - dipole coupling between orthogonal dipole moments . we applied the two methods to the two cases of atoms confined in a two - dimensional waveguide , and to atoms on a surface , in which either the polar or the azimuthal angle of the interatomic distance vector is known . in principle , these methods to determine the orientation can also be applied for the determination of both polar and azimuthal angle . the most promising ansatz is to make use of the resonance fluorescence intensity in a particular direction as discussed in sec . [ unknown - theta ] together with a rotation of the sample or of the driving laser polarization in order to fix one of the two angles at a value which renders the spectrum simpler ( e.g. , @xmath140 , @xmath173 or @xmath98 ) . then , the methods described in secs . [ unknown - phi ] and [ unknown - theta ] can be used to determine the other angle . the most straightforward implementation , however , strongly depends on the experimental possibilities to modify the setup . for example , in many cases , a rotation of the sample will be difficult . for many applications , the generalization to more than two particles is desirable . it remains to be seen whether methods based on the dipole - dipole interaction can also be applied in such cases . one approach could be to combine methods presented here together with a selective addressing of individual atoms at least in one or two dimensions , for example , by position - dependent state transfer . qg acknowledges support by the higher education commission ( hec ) , pakistan , the international max planck research school for quantum dynamics in physics , chemistry and biology , heidelberg , germany , and by deutscher akademischer austauschdienst ( daad ) . m. o. scully and k. drhl , phys . rev . a * 25 * , 2208 ( 1982 ) ; u. w. rathe and m. o. scully , lett . phys . * 34 * , 297 ( 1995 ) ; a. n. boto , p. kok , d. s. abrams , s. l. braunstein , c. p. williams , and j. p. dowling , phys . lett . * 85 * , 2733 ( 2000 ) ; m. dangelo , m. v. chekhova , and y. shih , phys . lett . * 87 * , 013602 ( 2001 ) . d. d. yavuz and n. a. proite , phys . a * 76 * , 041802(r ) ( 2007 ) ; m. kiffner , j. evers , and m. s. zubairy , phys . rev . lett . * 100 * , 073602 ( 2008 ) ; a. v. gorshkov , l. jiang , m. greiner , p. zoller , and m. d. lukin , phys . 100 * , 093005 ( 2008 ) . p. storey , m. collett , and d. walls , phys . lett . * 68 * , 472 ( 1992 ) ; r. quadt , m. collett , and d. f. walls , phys . rev . lett . * 74 * , 351 ( 1995 ) ; m. a. m. marte and p. zoller , appl . b * 54 * , 477 ( 1992 ) ; a. m. herkommer , h. j. carmichael , and w. p. schleich , quantum semiclass . opt . * 8 * , 189 ( 1996 ) ; f. l. kien , g. rempe , w. p. schleich , and m. s. zubairy , phys . rev . a * 56 * , 2972 ( 1997 ) ; e. paspalakis and p. l. knight , phys . a * 63 * , 065802 ( 2001 ) ; s. qamar , s .- y . zhu , and m. s. zubairy , phys . rev . a. * 61 * , 063806 ( 2000 ) ; d .- c . cheng , y .- p . niu , r .- x . li , and s .- q . gong , j. opt . b * 23 * , 2180 ( 2006 ) . m. macovei , j. evers , c. h. keitel , and m. s. zubairy , phys . a * 75 * , 033801 ( 2007 ) ; j. evers , s. qamar , and m. s. zubairy , phys . a * 75 * , 053809 ( 2007 ) s. qamar , j. evers , and m. s. zubairy , phys . a * 79 * , 043814 ( 2009 ) .	accurate measurement of relative distance and orientation of two nearby quantum particles is discussed . we are in particular interested in a realistic description requiring as little prior knowledge about the system as possible . thus , unlike in previous studies , we consider the case of an arbitrary relative orientation of the two atoms . for this , we model the atom with complete zeeman manifolds , and include parallel as well as orthogonal dipole - dipole couplings between all states of the two atoms . we find that it is possible to determine the distance of the two atoms independent of the orientation , as long as the particles are sufficiently close to each other . next , we discuss how in addition the alignment of the atoms can be measured . for this , we focus on the two cases of atoms in a two - dimensional waveguide and of atoms on a surface .
in this paper we are working on the energy critical nonlinear wave equation in 3 + 1 dimension in a domain @xmath0 where @xmath1 is a non - trapping obstacle with smooth boundary @xmath2 which enjoys the conservation of energy @xmath3 in the boundaryless case ( @xmath4 ) , the first results for the global existence were obtained by grillakis ( @xcite , @xcite ) . he showed that there are global smooth solutions of the critical wave equation , if the data is smooth . shatah and struwe ( @xcite , @xcite ) extended this theorem by showing that there are global solutions for the data lying in the energy space @xmath5 . they also obtained results for critical wave equation in higher dimensions . for the case of obstacles , the first results were due to smith and sogge @xcite . they showed that grillakis theorem extends to the case where @xmath6 is the complement of a smooth , compact , strictly convex obstacle with dirichlet boundary conditions . this result was later extended to the case of arbitrary domains in @xmath7 and data in the energy space by burq , lebeau and planchon @xcite . the case of the nonlinear critical neumann wave equation in 3-dimensions was subsequently handled by burq and planchon @xcite . more specifically , in this paper we are interested in asymptotics , i.e. how solutions to the nonlinear equation scatter to a solution to the homogeneous linear equation . in the boundaryless case ( @xmath8 ) , first results were obtained by bahouri - grard in @xcite ; in their paper they used the following decay estimate proved by bahouri - shatah @xcite @xmath9 to get @xmath10 and thus scattering . moreover , in @xcite bahouri - grard used profile decomposition to show that @xmath11 is also controlled by a universal function of the energy @xmath12.then the scattering result was extended to the case of star - shaped obstacles ( @xmath13 for @xmath14 with @xmath15 the outward pointing unit normal vector to @xmath16 ) by blair , smith , and sogge in @xcite . they used the same @xmath17 decay estimate proved by bahouri - shatah in @xcite after they extended it to their case of obstacles making slight modifications on the proof to handle the boundary term . in the papers by bahouri - shatah @xcite and blair , smith , and sogge @xcite , the @xmath17 decay estimate which is the main key to prove scattering was proved using the method of multipliers . the method of multipliers is also called friedrichs abc method as it dates back to kurt o. friedrichs in the 1950 s . the idea of this method is to multiply the equation with a factor @xmath18 , with @xmath19 is a linear first - order differential operator , defined as @xmath20 and then to express the product as a divergence or energy identity of the form @xmath21 and finally to integrate this divergence identity over a domain in @xmath22 and subsequently derive the required estimates . the only case where the differential multiplier is adapted to both the wave equation in terms of commutation ( avoiding remaining terms ) and the geometry of the obstacle in terms of the sign of the boundary term , is the star - shaped case . the method of multipliers was used in the 1960 s and 1970 s to prove uniform decay results for the homogeneous linear wave equation ( @xmath23 ) outside obstacles . cathleen s. morawetz was the first to succeed in proving uniform local energy decay for star - shaped obstacles with dirichlet boundary condition using this method ( @xcite and @xcite ) . since then , the results of morawetz have been considerably improved . better decay rates have been achieved ( as in odd dimensions @xmath24 , huygen s principle has been shown to imply an exponential rate of decay whenever there is some sort of decay @xcite , @xcite ) . moreover , the class of obstacles under consideration has been enlarged ; decay results have been derived for a special case of non trapping obstacles referred to as almost star - shaped regions(ivrii @xcite ) and for non trapping obstacles with simple and direct geometrical generalizations to the star - shaped such as the illuminated from interior(bloom and kazarinoff @xcite ) and the illuminated from exterior(bloom and kazarinoff @xcite , liu @xcite ) . for these cases , decay results have been proved using the method of multipliers after generalizing the multipliers to suit the geometry of the obstacle , and although these generalized multipliers lead to volume integrals that were avoided before , it turned out that these integrals were actually useful in the estimates . other wider generalizations that include all the above geometries later followed , strauss @xcite proved uniform local energy decay for the homogeneous linear wave equation in exterior domains in @xmath25 , provided a strictly expansive vector field ( now called the straussian vector field ) exists , that leaves @xmath26 strictly invariant , then these straussian vector fields were generalized by morawetz , ralston , and strauss @xcite by introducing escape functions , using them to construct a pseudo - differential operator @xmath27 , and finally setting @xmath28 as a multiplier . though the microlocal methods of morawetz , ralston , and strauss provided general results for the linear case , they can not be easily extended to the nonlinear wave equation for which the star - shapedness has been so far a restriction to obtain decay results . thus , we consider in this paper obstacles with explicit geometry that is a direct generalization of the star - shapedness , and we prove the @xmath17 decay estimate and thus the scattering for such obstacles using friedrichs abc method after generalizing the multipliers to suit our case . the price to pay for such a generalization is that the our multiplier , which is adapted to the geometry of the obstacle , no longer has exact commutation properties with the wave operator . therefore , unlike the star - shaped case we do get volume integrals that we deal with using a gronwall - like argument . our main results are : [ l6decay ] suppose @xmath29 solves the nonlinear wave equation and that @xmath1 is a non - trapping obstacle with regular boundary that can be illuminated from its exterior by a strictly convex body @xmath30 satisfying the geometric condition : @xmath31 where * @xmath32 is the algebraic distance from @xmath33 to @xmath14 along the exterior normal to @xmath33 . * @xmath34 where @xmath35 ( @xmath36 ) are the radii of curvature of @xmath33 ( @xmath37 ) . then @xmath38 we can construct obstacles that are illuminated from the exterior or from the interior that satisfy the condition . in particular , it would be easy to see this for illuminated from interior obstacles , where the illuminating body is inside the obstacle and thus @xmath39 , by considering a dog bone like obstacle ( figure [ fig : dogbone ] ) that is a small perturbation of the star - shaped . remark that if the data has compact support , the computation that proves the above result provides an explicit decay rate for the local energy . in particular , it recovers bloom - kazarinoff for the linear equation ( @xcite ) without using the fact that @xmath40 . as a result of this decay estimate , we get scattering [ scattering ] suppose @xmath29 solves the nonlinear wave equation and that @xmath1 is a non - trapping obstacle with regular boundary that can be illuminated from its exterior by a strictly convex body satisfying the geometric condition : @xmath41 then there exists unique solutions @xmath42 to the homogeneous linear problem @xmath43 such that @xmath44 moreover , @xmath29 satisfies : @xmath45 * acknowledgements. i would like to thank fabrice planchon for suggesting the problem and commenting on the manuscript . we consider in this paper illuminated from exterior obstacles , defined as such ( liu @xcite ) : [ exterior ] we say that the boundary of an exterior domain @xmath46 ( or the obstacle @xmath1 ) can be illuminated from the exterior if and only if there exists a convex body @xmath30 containing @xmath16 with smooth boundary @xmath33 such that @xmath16 is filled by a family of non - intersecting rays normal to @xmath33 . each ray is completely contained in @xmath6 in the following sense : for each @xmath47 there exists a unique @xmath48 and a number @xmath49 such that @xmath50 where @xmath51 is the outward unit normal to @xmath33 at @xmath52 , and @xmath53 this definition actually generalizes the following definition of illuminated from interior obstacles introduced by bloom and kazarinoff @xcite : [ illuminated ] we say that a body @xmath1 can be illuminated from the interior if and only if there exists a smooth convex body @xmath30 inside @xmath1 such that @xmath54 is filled by a family of non - intersecting rays normal to @xmath33 and such that each ray intersects @xmath16 exactly once . in fact , every body that can be illuminated from the interior can also be illuminated from its exterior by enlarging the original convex body ( liu @xcite ) ; thus our results proven for illuminated from exterior obstacles also hold for illuminated from interior . furthermore , as we mentioned in the introduction , these geometries are direct generalizations of the star - shaped . more precisely , the illuminated from interior is a generalization of the strict star - shapedness ( @xmath55 ) . the condition @xmath55 implies that each ray beginning at the origin intersects @xmath16 exactly once , which means that the interior of a strictly star - shaped obstacle can be illuminated by a source of light situated at the origin . a small ball centered at the origin is contained in the interior of @xmath1 and the above light rays are perpendicular to the surface of this ball , hence our strictly star - shaped is illuminated from its interior by this ball.an example of a non star - shaped body that can be illuminated from interior is a dog bone(figure [ fig : dogbone ] ) , and a snake - shapedbody ( figure [ fig : snake ] ) is an example of an obstacle that can not be illuminated from its interior but can be illuminated from the exterior . in this section we will introduce the coordinate system that we are going to use which is the one in the paper by liu @xcite in which he was dealing with similar obstacle for the linear equation . we again denote by @xmath33 the smooth and convex surface of the illuminating body . let @xmath56 be cartesian coordinates in @xmath57 with the origin inside @xmath30 and @xmath1 . if @xmath58 is on @xmath33 , then in a neighborhood of @xmath58 we choose the parametric curves to be the two principal curves on @xmath33 . if the neighborhood of @xmath58 is an all - umbilic surface , then we still can choose the parametric curves to be orthogonal to each other . furthermore , we let the parameters be the arc - length parameters . thus , if @xmath59 , then @xmath58 is given in local coordinates by @xmath60 where @xmath61 and @xmath62 are the parameterizations of the arc - length of the principal curves near @xmath58 . a finite number of @xmath63 coordinate patches cover @xmath33 . next , corresponding to each point @xmath64 on @xmath33 , we make the choice @xmath65 , where @xmath66 where @xmath67 is the unit exterior normal to @xmath33 , with @xmath68 . by definition [ exterior ] , for each @xmath69 , there is a unique triple @xmath70 with @xmath71 such that @xmath72 we denote by @xmath73 and @xmath74 the principal curvatures at @xmath64 and @xmath75 the radii of curvature of @xmath33 . we assume @xmath76 ( @xmath77 ) . furthermore , we always assume that @xmath78 this condition implies that for every @xmath79 , we have @xmath80 ( @xmath36 ) since @xmath81 with @xmath82 , where @xmath32 corresponds to the point on @xmath16 associated with @xmath58 . generically , definition [ exterior ] implies the condition ( @xcite page 26 lemma 2.1 , @xcite page 316 remark after lemma 1 ) . moreover , note that for a star - shaped obstacle , which is in fact an obstacle that is illuminated from the exterior by some ball @xmath83 , @xmath84 is nothing but @xmath85 . this explains the significance of this value and makes the need of such an assumption in a computation that is a generalization of the star - shaped totally logical . now , we state the following geometrical lemma that we will use later : [ a0 ] there exist a constant @xmath86 such that @xmath87 . the existence of @xmath88 is due to the boundedness of the the obstacle @xmath1 and the illuminating body @xmath30 . in fact , if @xmath89 ( @xmath90 ) then @xmath91 where @xmath92 and if @xmath93 ( @xmath94 ) then @xmath95 for some positive constant @xmath96 . we also recall the following lemmas about the coordinate system , these lemmas were originally stated and proved by bloom and kazarinoff @xcite for illuminated from interior obstacles , and then they were extended by liu @xcite for illuminated from exterior obstacles . [ lemma1 ] the level surfaces @xmath97 , @xmath98 @xmath99 define a set of local coordinate systems in @xmath26 with each ray @xmath100 normally incident on @xmath33 and @xmath101 where @xmath102 . [ lemma2 ] @xmath103 on @xmath16 . [ expressions ] we recall the following calculus formulas that we will use in our computation @xmath99 : @xmath104 and thus @xmath105 moreover , remark that for any scalar function @xmath106 and every vector field written in the new coordinate system @xmath107 we can express the gradient and the divergence as follows : @xmath108 and @xmath109\ ] ] in particular , we have @xmath110 @xmath111 @xmath112 denote by @xmath113 and @xmath114 thus @xmath115 we must show that for any @xmath116 , there exists @xmath117 such that whenever @xmath118 , @xmath119 first , we begin by multiplying the wave equation @xmath120 by @xmath121 , we get the following divergence or energy identity @xmath122 where @xmath123 denotes the energy density . integrating over the region @xmath124 , where @xmath125 and @xmath126 is a positive constant chosen such that @xmath127 and the illuminating body @xmath128 , and using the divergence theorem and the dirichlet boundary condition , we get @xmath129 where the flux on the mantle is defined by : @xmath130 with @xmath131 since the solution has finite energy , we may select @xmath126 large so that the right hand side of is less than @xmath132 . hence , it will suffice to show the existence of @xmath117 such that whenever @xmath133 we have @xmath134 this is a consequence of the following proposition : [ differentialinequality ] suppose @xmath29 solves the nonlinear wave equation and that @xmath1 is a non - trapping obstacle with regular boundary that can be illuminated from its exterior by a strictly convex body satisfying the geometric condition : @xmath41 then @xmath135 for some arbitrary @xmath136 and where @xmath137 and all the other constants depend on the geometry of the obstacle . as a result of proposition [ differentialinequality ] , we will show that given @xmath116 , @xmath138 such that @xmath139 we have @xmath140 for simplicity , let : @xmath141 @xmath142 @xmath143 with @xmath144 . thus we want to show the existence of @xmath117 such that @xmath139 we have @xmath145.@xmath146 from the differential inequality in proposition [ differentialinequality ] we have @xmath147 so @xmath148 @xmath149 @xmath150 @xmath151 hence , @xmath152 with @xmath153 but , note that @xmath154{}0 $ ] by the classical energy - conservation law on the exterior of a truncated cone stated above . thus , @xmath155 such that @xmath156 we have @xmath157 and choose @xmath158 such that @xmath159 hence , @xmath160 and @xmath161 is bounded : @xmath162 moreover , @xmath163 thus , for all @xmath133 , @xmath164 and by , @xmath145 which is . now , from proposition [ differentialinequality ] , we have @xmath165 hence , for all @xmath166 , we have @xmath167 which ends the proof of theorem [ l6decay ] . the method we are going to use to prove our result is the method of multipliers and we will generalize the morawetz multipliers that were used for star - shaped obstacles in way that suits our obstacle . we multiply the wave equation by @xmath168 where @xmath127 is the positive constant chosen in the previous section and @xmath169 is a vector field defined as follows : @xmath170 and we get the following divergence identity : @xmath171 where @xmath172 where @xmath173.integrating the divergence identity over the truncated cone @xmath174 and applying the divergence theorem , we get @xmath175 where @xmath176 denotes the lebesgue measure on the corresponding surface and @xmath15 is the outward pointing unit normal vector to @xmath16 ; and where @xmath177 and @xmath178 now , we deal with the terms of the integral equality in order to get the desired differential inequality.the boundary termby the dirichlet boundary condition , we get : @xmath179 since @xmath180 with @xmath181 ( by our assumption ) and @xmath103 ( lemma [ lemma2 ] ) . * the terms on the time slices*we have @xmath182 introduce @xmath183^{2}+\frac{1}{4}(t_{2}+m-(s+\rho_{2m}))\left[\partial_{t}u-\frac{\partial_{s}((s+\rho_{2m})u)}{{s+\rho_{2m}}}\right]^{2}\\ & \;\;\;\;+ ( t_{2}+m)\frac{u^{6}}{6}\\ & = ( t_{2}+m)\left(\frac{\|\partial_{t}u\|^{2}}{2}+\frac{\|\partial_{s}u\|^{2}}{2}+\frac{u^{6}}{6}\right)+(\partial_{t}u)(s+\rho_{2m})\partial_{s}u+(\partial_{t}u)u+\frac{t_{2}+m}{2}\left(\frac{u^{2}}{(s+\rho_{2m})^{2}}+\frac{2u\partial_{s}u}{s+\rho_{2m}}\right)\end{aligned}\ ] ] by remark [ expressions ] , we have @xmath184 where @xmath185 and @xmath186 so @xmath187 thus , @xmath188 now , integrating by parts and using dirichlet boundary condition , we get @xmath189 where @xmath190 is the measure on @xmath191 and @xmath192 but @xmath193 so @xmath194 similarly we get @xmath195 * the term on the mantle*on the mantle @xmath196 , we have @xmath197 , and recall that @xmath198 ( remark [ expressions ] ) , thus we get : @xmath199 so @xmath200 now setting @xmath201 on the mantle , we have @xmath202 the second term in can be written as such @xmath203 where @xmath204 integrating by parts and using the dirichlet boundary condition , we get @xmath205 thus the term on the mantle becomes @xmath206 the remainder term * we have @xmath207 thus by remark [ expressions ] , we get @xmath208 since @xmath80 and @xmath209 then @xmath210 and @xmath211 now , in the following remainder term @xmath212 we have @xmath213 imposing the following geometric condition @xmath214 we get that ( recall that @xmath215 ) @xmath216 and thus @xmath217 we still have to deal with the following term in @xmath218 : @xmath219 we have @xmath220 ; and by remark [ expressions ] , we have @xmath221 @xmath222 @xmath223 hence @xmath224 using @xmath225 we get @xmath226 moreover , we have @xmath227 hence @xmath228 recall that @xmath229 and note that @xmath230 hence @xmath231 substituting this in , we get : @xmath232 @xmath233 integrating by parts the last term and using the dirichlet boundary condition , we get : @xmath234 and we have @xmath235 hence , @xmath236 now substituting in we get @xmath237 and note that since we imposed the geometric condition @xmath214 we have @xmath238 and thus the last term in is nonnegative . * * the differential inequality**now , summing up all the terms ( , , , , , , and ) in the integral equality and dropping the nonnegative terms , we get the following differential inequality : @xmath239 recall that for @xmath36 we have : @xmath240^{2}+\frac{1}{4}(t_{i}+m-(s+\rho_{2m}))\left[\partial_{t}u-\frac{\partial_{s}((s+\rho_{2m})u)}{s+\rho_{2m}}\right]^{2}\\ & \;\;\;\;+(t_{i}+m)\frac{u^{6}}{6}\end{aligned}\ ] ] thus @xmath241 we also have , @xmath242 so we get @xmath243 we have @xmath244 and on @xmath245 we have : @xmath246 , thus @xmath247 moreover , by the geometric condition we imposed @xmath214 we have @xmath248 thus @xmath249 so @xmath250 recall that there exist a constant @xmath86 such that @xmath87 ( lemma [ a0 ] ) . thus , @xmath251 and by hardy s inequality : @xmath252 we get that @xmath253 so @xmath254 where @xmath255 and @xmath256 are constants that depend on the geometry of the obstacle and @xmath257 is the conserved energy . now , the term on the mantle @xmath258 can be written as follows : @xmath259 similarly to , we have @xmath260 hence @xmath261 where @xmath262 and @xmath263 are constants that depend on the geometry of the obstacle , and where @xmath264 @xmath265 thus , @xmath266 thus becomes @xmath267 split @xmath6 into @xmath268 and @xmath269 , where @xmath270 is a constant that depends on the geometry of the obstacle to be later specified , then the last term in becomes @xmath271 @xmath272 with @xmath273 . since @xmath87 and using hardy s inequality , we get @xmath274 hence , @xmath275 where @xmath276 and @xmath277 are constants that depend on the geometry of the obstacle . thus @xmath278 with @xmath279 the geometric condition @xmath214 which we assumed so far implies that @xmath280 which is not enough as we want @xmath281 for the proof of the @xmath17 decay estimate ( theorem [ l6decay ] ) . thus , we impose a stronger condition @xmath41 on the other hand , @xmath282 using and , becomes @xmath283 now , setting @xmath284 and @xmath285 for some @xmath136 and choosing @xmath270 such that @xmath286 , yields @xmath287 which ends the proof of proposition [ differentialinequality ] . as a result of the @xmath17 decay estimate we get the scattering result of corollary [ scattering ] . the proof of this corollary was done in the paper of blair , smith , and sogge @xcite and we replicate it here for the sake of completeness.we have the following strichartz estimate on functions @xmath288 satisfying homogeneous dirichlet boundary condition on non - trapping obstacles @xmath289 and we define the conserved energy of the linear equation @xmath290 attention was restricted to the @xmath291 function , as symmetric arguments will yield the existence of a @xmath292 asymptotic to @xmath29 at @xmath293 . as remarked in @xcite , it is enough to prove as follows as a consequence . they first established the existence of the wave operator , that is for any solution @xmath294 to the linear equation , there exists a unique solution @xmath29 to the non linear equation such that @xmath295 given , for any @xmath296 one may select @xmath297 large so that @xmath298 . given any @xmath288 satisfying @xmath299 , we have a unique solution to the linear problem @xmath300 @xmath301 as the right hand side is in @xmath302 . the estimate then also ensures that @xmath303 hence for @xmath304 sufficiently small , the map @xmath305 is seen to be a contraction on the ball of radius @xmath304 in @xmath306 . the unique fixed point @xmath307 can be uniquely extended over all of @xmath308 . hence taking @xmath309 shows the existence of the wave operator . to see that the wave operator is surjective , they used the @xmath310 decay estimate which we proved in theorem [ l6decay ] for our obstacle . this decay estimate establishes that the non linear effects of the solution map for the non linear equation diminish as time evolves . by the result of theorem [ l6decay ] , given any @xmath311 , there exists @xmath297 sufficiently large such that @xmath312 hence for any @xmath313 we obtain the following for any solution @xmath29 to @xmath314;l^{10}(\omega))}+\left\\|u\right\\|_{l^{4}([t , s];l^{12}(\omega))}&\leq c\left(e+\left\\|u^{5}\right\\|_{l^{1}([t , s];l^{2}(\omega))}\right)\\ & \leq ce+c\epsilon\left\\|u\right\\|^{4}_{l^{4}([t , s];l^{12}(\omega))}\end{aligned}\ ] ] a continuity argument now yields @xmath315 and by time reflection argument , follows . however , this implies that the linear problem @xmath316 @xmath317 admits a solution , showing that the wave operator is indeed surjective as @xmath318 is the desired solution to .	scattering for the energy critical non - linear wave equation for domains exterior to non trapping obstacles in 3 + 1 dimension is known for the star - shaped case . in this paper , we extend the scattering for a class of non star - shaped obstacles called illuminated from exterior . the main tool we use is the method of multipliers with weights that generalize the morawetz multiplier to suit the geometry of the obstacle .
our model problem consists of a system of chemicals that are coupled only through the reaction terms and diffuse independently of each other . given an integer @xmath0 , let @xmath1 be an ( @xmath2 ) vector of concentrations of chemical species , with @xmath3 , the spatial variable and @xmath4 , \ t>0,$ ] the time variable . the model we shall consider is of the following form ( see @xcite for details of the derivation ) : find @xmath5 , functions from @xmath6 into @xmath7 , such that for @xmath8 , @xmath5 satisfies @xmath9({\ensuremath{\boldsymbol{x}}},t)=f_{i}\left({\ensuremath{\boldsymbol{u}}}({\ensuremath{\boldsymbol{x}}},t)\right ) , & { \ensuremath{\boldsymbol{x}}}\in{\ensuremath{{\omega_t}}},t\in(0,t],\\ [ { \ensuremath{{{\ensuremath{\boldsymbol{\nu}}}}}}\cdot\nabla{u}_i]({\ensuremath{\boldsymbol{x}}},t)=0 , & \text { } { \ensuremath{\boldsymbol{x}}}\in\partial{\ensuremath{{\omega_t } } } , t>0,\\ u_i({\ensuremath{\boldsymbol{x}}},0)=u_i^{0}({\ensuremath{\boldsymbol{x } } } ) , & \text { } { \ensuremath{\boldsymbol{x}}}\in{\ensuremath{{\omega}}}_0 , \end{cases } \end{split}\ ] ] where @xmath10 is a simply connected bounded continuously deforming domain with respect to @xmath11 , with lipschitz boundary @xmath12 at time @xmath4 $ ] . the vector of nonlinear coupling terms @xmath13 is assumed to be locally lipschitz - continuous , @xmath14 is a vector of strictly positive diffusion coefficients , @xmath15 is a flow velocity generated by the evolution of the domain and the initial data @xmath16 is a bounded vector valued function . systems of this form arise in the theory of biological pattern formation @xcite . let @xmath17 be a simply connected time - independent reference domain with lipschitz boundary . we assume there exists a time - differentiable family of @xmath18-diffeomorphisms @xmath19 \rightarrow\omega_t$ ] such that at each instant @xmath20 $ ] and for each @xmath21 there exists a @xmath22 such that @xmath23 based on the derivation presented in @xcite we introduce a _ weak formulation _ associated with problem ( [ c : derivation : eqn : model_problem ] ) on the _ time - independent reference domain_. the problem is to find @xmath24 with @xmath25 such that , @xmath26 here here we state a theorem , and the associated assumptions under which the theorem holds , that shows the error in the _ semidiscrete _ scheme can be bounded by a computable a posteriori error estimator , based on the element _ residual_. our strategy to derive an a posteriori error estimate is similar to that employed in @xcite . we use energy arguments to show the residual is an upper bound for the error and the analysis is similar to the a priori case we have considered elsewhere @xcite . for the details of the proofs we refer to @xcite . we start by stating the semidiscrete scheme , find @xmath27\rightarrow{\ensuremath{\hat{\mathbb{v}}}}$ ] , such that for @xmath8 , @xmath28 } } } { \text { and } t\in(0,t ] } } , \\ \hat{u}^h_i(0)&=\lambda^h\hat{u}_i^0 , \end{aligned } \end{cases}\ ] ] where @xmath29 is a standard fe space made up of piecewise polynomial functions and @xmath30 is the lagrange interpolant . [ assum : mvt ] we assume that @xmath31 note this assumption is satisfied if we assume a global smallness condition on the mesh - size @xcite and that the continuous problem is well posed @xcite . we define the error in the semidiscrete scheme @xmath32.\ ] ] [ assum : energy_l2_dominance ] since we are primarily interested in problems posed on long time intervals , we wish to circumvent the use of gronwall s inequality . to this end we assume that the error in the @xmath33 norm converges faster than the error in the @xmath34 norm . we assume there exists @xmath35 and @xmath36 $ ] independent of the mesh - size @xmath37 such that @xmath38 thus @xmath39 where we have used the equivalence of norms between the reference and evolving domains . we note assumptions of this type have been used previously in @xcite and @xcite to obtain a posteriori estimates for quasilinear reaction - diffusion and nonlinear convection - diffusion problems . we start by introducing the _ residual _ @xmath40 ( the dual of @xmath41 ) a.e . in @xmath42 $ ] which satisfies @xmath43 where @xmath44 denotes the duality pairing between @xmath45 and its dual . we now show the residual is an upper bound for the error . suppose assumptions [ assum : mvt ] and [ assum : energy_l2_dominance ] hold . let @xmath46 satisfy ( [ c : fem_apriori : eqn : cont_weak_form_reference_ddtju ] ) and let the error @xmath47 and the residual @xmath48 be as in ( [ c : fem_apriori : eqn : error_defn ] ) and ( [ c : apriori_fem : eqn : residual_definition ] ) respectively . if the mesh - size satisfies a smallness condition ( see @xcite for details ) , then @xmath49 we now introduce a concrete error estimator . for simplicity we restrict the discussion to the case of @xmath50 elements and regular triangulations , the results may be straightforwardly generalised to higher order elements . for any simplex @xmath51 of the triangulation @xmath52 we denote by @xmath53 the of @xmath51 . let @xmath54 be the set of three edges of @xmath51 . let @xmath55 be an edge on the interior of @xmath17 , with outward pointing ( with respect to @xmath51 ) normal @xmath56 . we denote by @xmath57 the jump of @xmath58 across the edge @xmath55 . for boundary edges we take @xmath59 . the local error indicator is given by @xmath60 let @xmath61 and @xmath62 , @xmath8 be defined by ( [ c : apriori_fem : eqn : residual_definition ] ) and ( [ c : fem_apriori : eqn : estimator_defn ] ) respectively . there exists a @xmath63 that depends only on the shape regularity of the triangulation @xmath52 such that for @xmath8 , @xmath64;{\ensuremath{\operatorname{h}^{1}}}{({\ensuremath{{\hat{\omega } } } } ) } ) } } } { { \ensuremath{\operatorname{l}_{2}}}([0,t];{\ensuremath{\operatorname{h}^{1}}}{({\ensuremath{{\hat{\omega } } } } ) } ) } } .\ ] ] to complete the bound of the error by the estimator , we make an assumption about the error in the approximation of the initial data . [ assum : initial_l2_dominance ] we assume that the initial error in the @xmath65 norm converges faster than the error in the @xmath34 . we assume there exists @xmath66 and @xmath36 $ ] both independent of the mesh - size @xmath37 such that @xmath67 [ c : fem_apriori : thm : aposteriori ] let assumptions [ assum : mvt ] , [ assum : energy_l2_dominance ] and [ assum : initial_l2_dominance ] hold . let the error @xmath47 and the estimator @xmath68 @xmath8 be defined by ( [ c : fem_apriori : eqn : error_defn ] ) and ( [ c : fem_apriori : eqn : estimator_defn ] ) respectively . if the mesh - size is sufficiently small , for some @xmath63 , we have @xmath69 since the estimator @xmath70 is an upper bound for the error , we use it to drive a space - adaptive scheme . to ensure the efficiency of the adaptive scheme , we would have to show the estimator was also a lower bound for the error and we leave this extension for future work . here we reconcile our theoretical results with numerical computations . we start by presenting a time discretisation of the _ semidiscrete _ scheme , we discretise in time using a modified implicit euler method @xcite , in which the reaction terms are treated semi - implicitly while the diffusive terms are treated fully implicitly : find @xmath71 , such that for @xmath8 , @xmath72 , @xmath73^n,\hat{\phi}^n\right\rangle}_{{\ensuremath{{\hat{\omega}}}}}}+&{\ensuremath{\left\langle{d}_i[j{\ensuremath{\boldsymbol{k}}}{\nabla}\hat({u}^h_i)]^n,[{\ensuremath{\boldsymbol{k}}}\nabla\hat{\phi]^n}\right\rangle}_{{\ensuremath{{\hat{\omega}}}}}}\\ & = { \ensuremath{\left\langlej^n\widetilde{f}_i((\hat{u}^h_i)^n,(\hat{u}^h_i)^{n-1}),\hat{\phi}^n\right\rangle}_{{\ensuremath{{\hat{\omega } } } } } } { \ensuremath{\quad \forall}}\hat{\phi } \in { \ensuremath{\hat{\mathbb{v}}}}^n,\\ ( \hat{u}^h_i)^0&=\lambda^h\hat{u}_i^0 . \end{aligned } \end{cases}\ ] ] the adaptive algorithm we consider is based on the equidistribution marking strategy ( * ? ? ? * alg . 1.19 , pg . 45 ) , where elements are marked for refinement and coarsening with the goal of equidistributing the estimator value over all mesh elements . the marking strategy takes two parameters : the tolerance of the adaptive algorithm @xmath74 and a parameter @xmath75 . at each timestep elements are marked for refinement according to the following algorithm : start with @xmath77 the initial triangulation at time @xmath78 , tolerance @xmath74 and parameter @xmath79 @xmath80 solve the discrete linear problem on the mesh @xmath81 compute global error estimator @xmath82 and local error indicators @xmath83 mark @xmath51 for refinement adapt mesh @xmath81 to give @xmath84 @xmath85 solve the discrete linear problem on the mesh @xmath81 compute global error estimator @xmath82 and local error indicators @xmath83 elements are marked for coarsening in a similar way to the above , the difference being that if the local error indicator plus a coarsening indicator is less than a given tolerance on an element then the element is marked for coarsening ( * ? ? ? 48 ) . for numerical testing , we consider problem ( [ c : derivation : eqn : model_problem ] ) equipped with the schnakenberg kinetics @xcite : @xmath86 . the details of the implementation of the scheme are described elsewhere @xcite . we consider a simulation of an rds equipped with the schnakenberg kinetics , with parameter values @xmath87 , @xmath88 , @xmath89 , @xmath90 , adding source term such that the exact solution is known , on a domain with evolution of the form @xmath91 ^ 2,t\in[0,1].\ ] ] we used a sufficiently small timestep such that the error due to the time discretisation is negligible . the estimator values and eoc for a series of refinements is plotted in figure [ c : fem_apriori : fig : residual_estimator ] and we observe an eoc of 1 , as expected for @xmath50 elements providing numerical evidence for theorem [ c : fem_apriori : thm : aposteriori ] . we next present results for the schnakenberg kinetics , with parameter values @xmath92 , @xmath88 , @xmath89 , @xmath93 where no exact solution is known on a domain with evolution of the form @xmath94 ^ 2\text { and } t\in[0,1000].\ ] ] we consider an adaptive scheme based on the equidistribution marking strategy with parameters @xmath95 , @xmath96 and a fixed timestep of @xmath97 . figures [ c : numerics : fig : periodic_dof ] and [ c : numerics : fig : sinusoidal_adaptivity ] show the evolution of the degrees of freedom ( dofs ) and the change in discrete solution and snapshots of the activator profiles ( on the reference domain ) respectively . the number of dofs appears positively correlated with the domain size . the mesh is also well refined around the patterns during the evolution , illustrating the benefits of adaptive mesh refinement . we finish with an application to the case where the evolving domain is an _ evolving surface _ embedded in @xmath98 that is diffeomorphic to a time - independent planar domain . we have derived the model equations and corresponding finite element method elsewhere @xcite and thus only briefly state the details . the model for an rds posed on an evolving surface is of the form : find @xmath5 , functions from @xmath99 into @xmath7 , such that for @xmath8 , @xmath5 satisfies @xmath100({\ensuremath{\boldsymbol{x}}},t)- d_i{\ensuremath{\updelta}}_{\gamma_t}{u}_i({\ensuremath{\boldsymbol{x}}},t)=f_{i}\left({\ensuremath{\boldsymbol{u}}}({\ensuremath{\boldsymbol{x}}},t)\right ) , & { \ensuremath{\boldsymbol{x}}}\in\gamma_t , t\in(0,t],\\ [ { \ensuremath{{{\ensuremath{\boldsymbol{\nu}}}}}}\cdot\nabla_{\gamma_t}{u}_i]({\ensuremath{\boldsymbol{x}}},t)=0 , & \text { } { \ensuremath{\boldsymbol{x}}}\in\partial\gamma_t , t>0,\\ u_i({\ensuremath{\boldsymbol{x}}},0)=u_i^{0}({\ensuremath{\boldsymbol{x } } } ) , & \text { } { \ensuremath{\boldsymbol{x}}}\in\gamma_0 , \end{cases } \end{split}\ ] ] here the cartesian gradient and laplacian that appear in ( [ c : derivation : eqn : model_problem ] ) are replaced by the surface ( tangential ) gradient and laplace - beltrami operator . we assume the surface @xmath99 admits an _ orthogonal parameterisation _ to a planar domain which we denote by @xmath101\to\gamma_t\subset{\ensuremath{{\mathbb{r}}}}^3.\ ] ] under similar assumptions to those made in the planar case ( see @xcite for details ) the corresponding weak formulation on the reference domain @xmath17 is given by ( [ c : fem_apriori : eqn : cont_weak_form_reference_ddtju ] ) where the matrix @xmath102 and the determinant of the jacobian @xmath103 are given by @xmath104\quad \text{and } \quad j={\ensuremath{\left\vert}}\partial_{1}{\ensuremath{{\ensuremath{\boldsymbol{\mathcal{a}}}}}}{\ensuremath{\right\vert}}{\ensuremath{\left\vert}}\partial_{2}{\ensuremath{{\ensuremath{\boldsymbol{\mathcal{a}}}}}}{\ensuremath{\right\vert}}.\end{aligned}\ ] ] we consider an example with the schnakenberg kinetics ( [ eqn : schnakenberg_rds ] ) , with parameter values @xmath92 , @xmath88 , @xmath89 , @xmath105 where no exact solution is known on a domain with evolution of the form @xmath106 ^ 2,t\in[0,500].\ ] ] we once again consider an adaptive scheme based on the equidistribution marking strategy with parameters @xmath95 , @xmath107 and a fixed timestep of @xmath97 . figure [ fig : surface_adaptivity ] shows snapshots of the activator profiles ( on the surface and on the reference domain ) and the mesh of the reference domain . as the surface evolves , we observe the emergence of a large number of spots with small radii in the top left and bottom right hand corners of the domain ( where curvature is large and growth is fastest ) with annihilation of these spots as the domain contracts . the results clearly illustrate the influence of growth and curvature on pattern formation . the adaptive scheme appears to resolve the solution profiles and the mesh is well refined around the spots on the reference domain , capturing both the small radii spots that develop in the northwest and southeast corners and the large radii spots that develop elsewhere . finally , we remark that we have also considered space - time adaptive schemes based on an heuristic error indicator for the time adaptivity which appear to give dramatic improvements in efficiency @xcite . c venkataraman , o lakkis , and a madzvamuse . adaptive finite elements for semilinear reaction - diffusion systems on growing domains . in _ numerical mathematics and advanced applications 2011 : proceedings of enumath 2011 , the 9th european conference on numerical mathematics and advanced applications , leicester , september 2011 _ , page 71 . springer , 2013 .	we propose an adaptive finite element method to approximate the solutions to reaction - diffusion systems on time - dependent domains and surfaces . we derive a computable error estimator that provides an upper bound for the error in the semidiscrete ( space ) scheme . we reconcile our theoretical results with benchmark computations .
the importance of data cleaning and quality technologies for business practices is well recognized . data cleaning has been an active research topic in several communities including statistics , machine learning and data management . the quality of data suffers from typing mistakes , lack of standards for recording database fields , integrity constraints that are not enforced , inconsistent data mappings , etc . for years , data quality technology has grown independently from core data management . data quality tools became part of extract transform load ( etl ) technologies , commonly applied during the initial loading phase of data into a warehouse . although this might be a viable approach for data analytics , where data processed are static , it is far from acceptable for operational databases . dynamic databases however , face proliferating quality problems , that degrade common business practices . recently , there has been a major focus on tighter integration of data quality technology with database technology . in particular there has been research work on the efficient realization of popular data cleaning algorithms inside database engines as well as studies for the efficient realization of data quality primitives in a declarative way . the approaches are complementary , the former assuring great performance and the latter ease of deployment and integration with existing applications without modification of the underlying database engine . we are concerned with declarative implementations of data quality primitives in this thesis . in particular we study declarative realizations of several _ similarity predicates _ for the popular approximate ( flexible ) selection operation for data de - duplication @xcite . a similarity predicate @xmath0 is a predicate that numerically quantifies the similarity or closeness of two ( string ) tuples . given a relation @xmath1 , the approximate selection operation using similarity predicate @xmath0 , will report all tuples @xmath2 such that @xmath3 , where @xmath4 a specified numerical similarity threshold and @xmath5 a query tuple . approximate selections are special cases of the approximate join ( record linkage , similarity join ) operation @xcite . several efficient declarative implementations of this operation for specific similarity predicates have been proposed @xcite both for approximate selections and joins . in this thesis , we conduct a thorough study of declarative realizations of similarity predicates for approximate selections . we introduce and adapt novel predicates , realize them declaratively and compare them with existing ones for accuracy and performance . in particular we make the following contributions : * inspired by the success of _ tf - idf cosine similarity _ from information retrieval @xcite as a similarity predicate for approximate selections , we introduce declarative realizations of other successful predicates from information retrieval and in particular the popular bm25 measure . * we introduce declarative realizations of probabilistic similarity predicates inspired by language models from information retrieval @xcite and hidden markov models @xcite , suitably adapted for the case of approximate selections . * we present declarative realizations of previously proposed similarity predicates for the approximate selection problem and we propose a categorization of all measures both previously proposed and new according to their characteristics . * we present a thorough experimental study comparing all similarity predicates for accuracy and performance , under various types of quality problems in the underlying data . data quality has been an active research topic for many years . a collection of statistical techniques have been introduced initially for the record linkage problem @xcite . the bulk of early work on data quality was geared towards correcting problems in census files @xcite . a number of similarity predicates were developed taking into account the specific application domain ( i.e. , census files ) for assessing closeness between person names ( e.g. , jaro , jaro - winkler @xcite , etc ) . the work of cohen @xcite introduced the use of primitives from information retrieval ( namely cosine similarity , utilizing tf - idf@xcite ) to identify flexible matches among database tuples . a performance / accuracy study conducted by cohen et al . , @xcite demonstrated that such techniques outperform common predicates introduced for specific domains ( e.g. , jaro , jaro - winkler , etc ) . other techniques geared towards database tuples include the merge / purge technique @xcite . several predicates to quantify approximate match between strings have been utilized for dealing with quality problems , including edit distance and its variants @xcite . hybrid predicates combining notions of edit distance and cosine similarity have also been introduced @xcite . recently , @xcite presented ssjoin , a primitive operator for efficient set similarity joins . utilizing ideas from @xcite , such an operator can be used for approximate matching based on a number of similarity functions , including hamming distance , edit - distance and jaccard similarity . however , the choice of the similarity predicate in this approach is limited @xcite . the bulk of the techniques and predicates however have been introduced without a declarative framework in mind . thus , integrating them with applications utilizing databases in order to enable approximate selections is not very easy . gravano et al . @xcite , introduced a declarative methodology for realizing approximate joins and selections for edit distance . subsequently a declarative framework for realizing tf - idf cosine similarity was introduced @xcite . there has been a great deal of research in the information retrieval literature on weighting schemes beyond cosine similarity with tf - idf weighting . recent ir research has shown bm25 to be the most effective among the known weighting schemes @xcite . this weighting scheme models the distribution of within - document term frequency , document length and query term frequency very accurately . moreover , in the information retrieval literature , language modeling has been a very active research topic as an alternate scheme to weight documents for their relevance to user queries . starting with ponte and croft @xcite language models for information retrieval have been widely studied . hidden markov models ( hmm ) have been very successful in machine learning and they have been utilized for a variety of learning tasks such as named entity recognition and voice recognition@xcite . they have also been utilized for information retrieval as well @xcite . an experimental study on trec data demonstrated that an extremely simple realization of hmm outperforms standard tf - idf for information retrieval @xcite . several researchers @xcite have tried to formally reason about the relative goodness of information retrieval weighting schemes . let @xmath6 be a query string and @xmath7 a string record from a base relation @xmath8 . we denote by @xmath9 , @xmath10 the set of _ tokens _ in @xmath6 and @xmath7 respectively . we refer to substrings of a string as tokens in a generic sense . such tokens can be words or q - grams ( sequence of @xmath11 consecutive characters of a string ) for example . for @xmath6=`db lab ' , @xmath9=@xmath12`db ' , ` lab'@xmath13 for word - based tokenization and @xmath9=@xmath12`db ' , ` b l ' , ` la ' , ` lab'@xmath13 for tokenization using 3-grams . we refer to tokens throughout the thesis when referring to words or q - grams . we make the choice specific ( word or q - gram ) for techniques we present , when is absolutely required . in certain cases , we may associate a _ weight _ with each token . several weighting mechanisms exist . we present our techniques referring to weights of tokens , making the choice of the weighting scheme concrete when required . in chapter [ experimental ] we realize our techniques for specific choice of tokens and specific weighting mechanisms . our goal is to calculate a _ similarity score _ between @xmath6 and @xmath7 using a similarity predicate . we group similarity predicates into five classes based on their characteristics , namely : * * overlap predicates : * these are predicates that assess similarity based on the overlap of tokens in @xmath14 . * * aggregate weighted predicates : * predicates that assess similarity by manipulating weights ( scores ) assigned to elements of @xmath14 * * language modeling predicates : * predicates that are based on probabilistic models imposed on elements of @xmath14 * * edit based predicates : * predicates based on a set of edit operations applied between @xmath6 and @xmath7 . * * combination predicates : * predicates combining features from the classes above . the classes were defined by studying the properties of previously proposed similarity predicates as well as ones newly proposed herein . the first four classes encompass predicates introduced previously in various contexts for data cleaning tasks , with the exception of bm25 which to the best of our knowledge is the first time that is deployed for data cleaning purposes . the language modeling class of predicates draws from work on information retrieval and is introduced herein for data cleaning tasks . within each class we discuss declarative realizations of predicates . suppose @xmath9 is the set of tokens in the query string @xmath6 and @xmath10 is the set of tokens in the string tuple @xmath7 . the _ intersectsize _ predicate @xcite is simply the number of common tokens between @xmath6 and @xmath7 , i.e. : @xmath15 jaccard similarity @xcite is the fraction of tokens in @xmath6 and @xmath16 that are present in both , namely : @xmath17 if we assign a weight @xmath18 to each token @xmath19 , we can define weighted versions of the above predicates . _ @xcite is the total weight of common tokens in @xmath9 and @xmath10 , i.e. , @xmath20 . similarly , _ weightedjaccard _ is the sum of the weights of tokens in @xmath21 divided by the sum of the weights of tokens in @xmath22 . the predicates in this class encompass predicates widely adopted from information retrieval ( ir ) . a basic task in ir is , given a query , identifying _ relevant _ _ documents _ to that query . in our context , we would like to identify the _ tuples _ in a relation that are _ similar _ to a query string . given a query string @xmath6 and a string tuple @xmath7 , the similarity score of @xmath6 and @xmath7 in this class of predicates is of the form @xmath23 , where @xmath24 is the query - based weight of the token @xmath19 in string @xmath6 and @xmath25 is the tuple - based weight of the token @xmath19 in string @xmath7 . the tf - idf cosine similarity@xcite between a query string @xmath6 and a string tuple @xmath7 is defined as follows : @xmath26 where @xmath27 are the normalized tf - idf weights @xcite . the normalized tf - idf between a token @xmath19 and a string @xmath16 , @xmath28 is given by : @xmath29 the @xmath30 term makes the weight of a token inversely proportional to its frequency in the database ; the @xmath31 term makes it proportional to its frequency in @xmath16 . intuitively , this assigns low scores to frequent tokens and high scores to rare tokens . more discussion is available elsewhere @xcite . the @xmath32 similarity score between a query string @xmath6 and a tuple @xmath7 , is given as : @xmath33 where @xmath34 @xmath35 is a modified form of robertson - sparck jones weight : @xmath36 and @xmath37 is the number of tuples in the base relation @xmath1 , @xmath38 is the number of tuples in @xmath1 containing the token @xmath19 , @xmath39 is the frequency of occurrence of the token @xmath19 within tuple @xmath7 , @xmath40 is the number of tokens of tuple @xmath7 , @xmath41 is the average number of tokens per tuple , i.e. @xmath42 and @xmath43 , @xmath44 , and @xmath45 are independent parameters . for trec-4 experiments @xcite , @xmath46 $ ] , @xmath47 and @xmath48 $ ] . a language model , is a form of a probabilistic model . to realize things concretely , we base our discussion on a specific model introduced by ponte and croft @xcite . given a collection of documents , a language model is inferred for each ; then the probability of generating a given query according to each of these models is estimated and documents are ranked according to these probabilities . considering an approximate selection query , each tuple in the database is considered as a document ; a model is inferred for each tuple and the probability of generating the query given the model is the similarity between the query and the tuple . the similarity score between query @xmath6 and tuple @xmath7 is defined as : @xmath49 where @xmath50 is the probability of token @xmath19 occurring in tuple @xmath7 and is given as follows : @xmath51 @xmath52 is the maximum likelihood estimate of the probability of the token @xmath19 under the token distribution for tuple @xmath7 and is equal to @xmath53 where @xmath54 is raw term frequency and @xmath55 is the total number of tokens in tuple @xmath7 . @xmath56 is the mean probability of token @xmath19 in documents containing it , i.e. , @xmath57 where @xmath58 is the document frequency of token @xmath19 . this term is used since we only have a tuple sized sample from the distribution of @xmath59 , thus the maximum likelihood estimate is not reliable enough ; we need an estimate from a larger amount of data . the term @xmath60 is used to model the risk for a term @xmath19 in a document @xmath7 using a geometric distribution : @xmath61 @xmath62 is the expected term count for token @xmath19 in tuple @xmath7 if the token occurred at the average rate , i.e. , @xmath63 . the intuition behind this formula is that as the @xmath31 gets further away from the normalized mean , the mean probability becomes riskier to use as an estimate . finally , @xmath64 is the raw count of token @xmath19 in the collection , i.e. @xmath65 and @xmath66 is the raw collection size or the total number of tokens in the collection , i.e. @xmath67 . @xmath68 is used as the probability of observing a non - occurring token . the query generation process can be modeled by a discrete hidden markov process . figure [ hmm - fig ] shows a simple yet powerful two - state hmm for this process . the first state , labeled string " represents the choice of a token directly from the string . the second state , labeled general english " represents the choice of a token that is unrelated to the string , but occurs commonly in queries . suppose @xmath6 is the query string and @xmath7 is a string tuple from the base relation @xmath1 ; the similarity score between @xmath6 and @xmath7 , @xmath69 , is equal to the probability of generating @xmath6 given that @xmath7 is similar , that is : @xmath70 where : @xmath71 @xmath72 and @xmath73 and @xmath74 are transition probabilities of the hmm . the values for these parameters can be optimized to maximize accuracy given training data . an important and widely used class of string matching predicates is the class of edit - based predicates . in this class , the similarity between @xmath6 and @xmath7 is the transformation cost of string @xmath6 to @xmath7 , @xmath75 . more specifically @xmath75 is defined as the minimum cost sequence of edit operations that converts @xmath6 to @xmath7 . edit operations include _ copy , insert , substitute _ and _ delete _ characters in @xmath6 and @xmath7 @xcite . algorithms exist to compute @xmath75 in polynomial time @xcite but complexity is sensitive to the nature of operations and their operands ( individual characters , blocks of consecutive characters , etc ) . the edit similarity is therefore defined as : @xmath76 edit operations have an associated cost . in the levenstein edit - distance @xcite which we will refer to as edit - distance , the cost of copy operation is zero and all other operations have unit cost . other cost models are also possible @xcite . we present a general similarity predicate and refer to it as generalized edit similarity ( ges ) ( following @xcite ) . consider two strings @xmath6 and @xmath7 that are tokenized into _ word _ tokens and a weight function @xmath18 that assigns a weight to each word token @xmath19 . the transformation cost of string @xmath6 to @xmath7 , @xmath75 is the minimum cost of transforming @xmath6 to @xmath7 by a sequence of the following transformation operations : * _ token replacement _ : replacing word token @xmath77 in @xmath6 by word token @xmath78 in @xmath7 with cost @xmath79 \cdot w(t_1)$ ] , where @xmath80 is the edit similarity score between @xmath77 and @xmath78 . * _ token insertion _ : inserting a word token @xmath19 into @xmath6 with cost @xmath81 where @xmath82 , is a constant token insertion factor , with values between 0 and 1 . * _ token deletion _ : deleting a word token @xmath19 from @xmath6 with cost @xmath18 . suppose @xmath83 is the sum of weights of all word tokens in the string @xmath6 . we define the _ generalized edit similarity _ predicate between a query string @xmath6 and a tuple @xmath7 as follows : @xmath84 a related predicate is the softtfidf predicate @xcite . in softtfidf , normalized tf - idf weights of word tokens are used along with cosine similarity and any other similarity function @xmath85 to find the similarity between word tokens . therefore the similarity score , @xmath86 , is equal to : @xmath87 where @xmath88 are the normalized tf - idf weights and @xmath89 is the set of words @xmath90 such that there exists some @xmath91 such that @xmath92 . we now describe declarative realizations of predicates in each class . we present declarative statements using standard sql expressions . for all predicates , there is a preprocessing phase responsible for tokenizing strings in the base relation , @xmath1 , and calculating as well as storing related weight values which are subsequently utilized at query time . tokenization of relation @xmath1 ( ` base_table ` ) creates the table ` base_tokens ` ` ( tid , token ) ` , where @xmath93 is a unique token identifier for each tuple of ` base_table ` and _ token _ an associated token ( from the set of tokens corresponding to the tuple with identifier @xmath93 in ` base_table ` ) . the query string is also tokenized on the fly ( at query time ) creating the table ` query_tokens(token ) ` . in the rest of this chapter , we present sql expressions required for preprocessing and query time approximate selections for the different predicates . in some cases , we re - write formulas to make them amenable to more efficient declarative realization . the main sql codes are given along with their description here . appendix a contains detailed sql expressions . the intersectsize predicate requires token generation to be completed in a preprocessing step . sql statements to conduct such a tokenization , which is common to all predicates we discuss , is available in appendix a. the sql statement for approximate selections with the intersectsize predicate is shown on figure [ intersect ] . the jaccard coefficient predicate can be efficiently computed by storing the number of tokens for each tuple of the ` base_table ` during the preprocessing step . for this reason we create a table ` base_ddl(tid , token , len ) ` where ` len ` is the number of tokens in tuple with tuple - id ` tid ` . the sql statement for conducting approximate selections with the jaccard predicate is presented in figure [ jaccard ] . select & r1.tid , count ( * ) + from & base_tokens r1 , query_tokens r2 + where & r1.token = r2.token + group by & r1.tid + + select & s1.tid , count()/(s1.len+s2.len - count ( ) ) + from & base_ddl s1 , query_tokens r2 , + & ( select count ( * ) as len + & from query_tokens ) s2 + where & s1.token = r2.token + group by & s1.tid , s1.len , s2.len + the weighted overlap predicates require calculation and storage of the related weights for tokens of the base relation during preprocessing . for the weightedmatch predicate , we store during the preprocessing step the weight of each token redundantly with each ` tid , token ` pair in a table ` base_tokens_weights(tid , token , weight ) ` in order to avoid an extra join with a table ` base_weight(token , weight ) ` at query time . in order to calculate the similarity score at query time , we use sql statements similar to that used for the intersectsize predicate ( shown in figure [ intersect ] ) but replace table ` base_tokens ` by ` base_tokens_weights ` and ` count ( * ) ` , by ` sum(r1.weight ) ` . for the weightedjaccard predicate , we create during preprocessing a table ` base_ddl(tid , token , weight , len ) ` where ` weight ` is the weight of ` token ` and ` len ` is the sum of weights of tokens in the tuple with tuple - id ` tid ` . the sql statement for approximate selections using this predicate is the same as the one shown in figure [ jaccard ] but ` count ( * ) ` is replaced by ` sum(weight ) ` . the sql implementation of the tf - idf cosine similarity predicate has been presented in @xcite . during preprocessing , we store tf - idf weights for the base relation in relation ` base_weights(tid , token , weight ) ` . a weight table ` query_weights(token , weight ) ` for the query string is created on the fly at query time . the sql statements in figure [ awsql ] will calculate the similarity score for each tuple of the base table . select & r1w.tid , sum(r1w.weightr2w.weight ) + from & base_weights r1w , query_weights r2w + where & r1w.token = r2w.token + group by & r1w.tid + realization of bm25 in sql involves generation of the table ` base_weights(tid , token , weight ) ` storing the weights for tokens in each tuple of the base relation . these weights ( @xmath25 ) consist of two parts that could be considered as modified versions of tf and idf . for a complete set of sql statements implementing the required preprocessing , refer to appendix a. the query weights table ` query_weights(token , weight ) ` can be created on the fly using the following subquery : + ( select tf.token , tf.tf(@xmath44+1)/(@xmath44+tf.tf ) as weight + from ( select t.token , count ( * ) as tf + from query_tokens t + group by t.token ) tf ) + the sql statement shown in figure [ awsql ] will calculate bm25 similarity scores . in order to calculate language modeling scores efficiently , we rewrite the formulas and finally drop some terms that would not affect the overall accuracy of the metric . calculating the values in equations ( [ eqlm3 ] ) and ( [ eqlm4 ] ) is easy . we build the following relations during preprocessing : ` base_tf(tid , token , tf ) ` where ` tf`@xmath94 . + ` base_dl(tid , dl ) ` where ` dl`@xmath95 . + ` base_pml(tid , token , pml ) ` where ` pml`@xmath96 . + ` base_pavg(token , pavg ) ` where ` pavg`@xmath97 . + ` base_freq(tid , token , freq ) ` where ` freq`@xmath98 . + ` base_risk(tid , token , risk ) ` where ` risk`@xmath99 . we omit most of the sql statements in this chapter for readability . full sql statements are available in appendix a. in order to improve the performance of the associated sql queries , we rewrite the final score formula of equation ( [ eqlm1 ] ) , as follows : @xmath100 \times \frac{\displaystyle \prod_{\forall t}(1 - \hat{p}(t\|m_d ) ) } { \displaystyle \prod_{t \in \cal{q}}(1 - \hat{p}(t\|m_d ) ) } \end{aligned}\ ] ] we slightly change ( [ eqlm5 ] ) to the following : @xmath101 \times \frac{\displaystyle \prod_{\forall t \in \cal{d}}(1 - \hat{p}(t\|m_d ) ) } { \displaystyle \prod_{t \in \cal{q } \cap \cal{d}}(1 - \hat{p}(t\|m_d ) ) } \end{aligned}\ ] ] this change results in a large performance gain , since the computation is restricted to the tokens of the query and the tokens of a tuple ( as opposed to the entire set of tokens present in the base relation ) . experiments demonstrate that accuracy is not considerably affected . in equation ( [ eqlm2 ] ) , we only materialize the first part ( i.e. , values of tokens that are present in the tuple @xmath7 ) in the relation ` base_pm ` during preprocessing ( storing the second part would result in unnecessary waste of space ) . we therefore have to divide all formulas that use @xmath50 into two parts : one for tokens present in the tuple under consideration and one for all other tokens . so we rewrite the first term in equation ( [ eqlm6 ] ) as follows : @xmath102 the term @xmath103 in the above formula is constant for any specific query string , so it can be dropped , since the goal is to find most similar tuples by ranking them based on the similarity scores . therefore , equation ( [ eqlm6 ] ) can be written as follows : @xmath104 this transformation allows us to efficiently compute similar tuples by just storing @xmath105 and @xmath68 for each pair of @xmath19 and @xmath7 . thus , we create table ` base_pm(tid , token , pm , cfcs ) ` where ` pm ` @xmath106 and ` cfcs ` @xmath107 as the final result of the preprocessing step . we also calculate and store the term @xmath108 during preprocessing in relation ` base_sumcompbase(tid , sumcompm ) ` . the query - time sql statement to calculate similarity scores is shown in figure [ lm ] . the subquery in the statement computes the three terms in equation [ eqlm8 ] that include intersection of query and tuple tokens and therefore needs a join between the two token tables . the fourth term in the equation is read from the table stored during the preprocessing as described above . [ cols= " < " , ] to further investigate the effect of pruning based on idf on performance and accuracy of predicates , we examine the idf distribution in our data sets . figure [ idf - chart ] shows the distribution of idf weights for cu1 data set . the distribution is similar in all other data sets . as it can be seen , there is a huge number of tokens with low idf . for this data set , a pruning rate of @xmath109 will drop nearly 150,000 out of 250,000 tokens which results in a huge performance gain and a very little drop in accuracy of predicates ( except unweighted naive predicates that benefit from pruning as described above ) . we presented an exhaustive evaluation of approximate selection predicates by grouping them into five classes based on their characteristics : overlap predicates , aggregate weighted predicates , edit - based predicates , combination predicates and language modeling predicates . we experimentally show how predicates in each of these classes perform in terms of accuracy , preprocessing and execution time . within our framework , the overlap predicates are relatively efficient but have low accuracy . edit based predicates perform worse in terms of accuracy but are relatively fast due to the filtering step they employ . the aggregate weighted predicates , specifically bm25 , perform very well both in terms of accuracy and efficiency . both the predicates from the language modeling cluster perform well in terms of accuracy . moreover , hmm is as fast as simple overlap predicates . the combination predicates are considerably slow due to their two levels of tokenization . among the combination predicates , ges based predicates are robust in handling edit errors but fail considerably in capturing token swap errors . softtfidf with jaro - winkler performs nearly equal to bm25 and hmm and is among the best in terms of accuracy , although it is the slowest predicate . this establishes the effectiveness of bm25 and hmm predicates for approximate matching in large databases . we proposed new similarity predicates for approximate selections based on probabilistic information retrieval and presented their declarative instantiation . we presented an in - depth comparison of accuracy and performance of these new predicates along with existing predicates , grouping them into classes based on their primary characteristics . our experiments show that the new predicates are both effective as well as efficient for data cleaning applications . 10 r. ananthakrishna , s. chaudhuri , and v. ganti . eliminating fuzzy duplicates in data warehouses . in _ proceedings of the 28th international conference on very large databases ( vldb ) _ , 2002 . arvind arasu , venkatesh ganti , and raghav kaushik . efficient exact set - similarity joins . in _ proceedings of the 32nd international conference on very large data bases ( vldb ) _ , pages 918929 . vldb endowment , 2006 . andrei z. broder , moses charikar , alan m. frieze , and michael mitzenmacher . min - wise independent permutations . , 60(3):630659 , 2000 . surajit chaudhuri , kris ganjam , venkatesh ganti , and rajeev motwani . robust and efficient fuzzy match for online data cleaning . in _ sigmod03 : proceedings of the 2003 acm sigmod international conference on management of data _ , pages 313324 , june 2003 . surajit chaudhuri , venkatesh ganti , and raghav kaushik . a primitive operator for similarity joins in data cleaning . in _ icde 06 : proceedings of the 22nd international conference on data engineering ( icde ) _ , page 5 , washington , dc , usa , 2006 . ieee computer society . william w. cohen . integration of heterogeneous databases without common domains using queries based on textual similarity . in _ sigmod 98 : proceedings of the 1998 acm sigmod international conference on management of data _ , pages 201212 , new york , ny , usa , 1998 . acm press . william w. cohen , pradeep ravikumar , and stephen e. fienberg . a comparison of string distance metrics for name - matching tasks . in _ proceedings of ijcai-03 workshop on information integration on the web ( iiweb-03 ) _ , pages 7378 , 2003 . j. b. copas and f. j. hilton . record linkage : statistical models for matching computer records . , pages 287320 , 1990 . ivan p. fellegi and alan b. sunter . a theory for record linkage . , 64(328):11831210 , 1969 . helena galhardas , daniela florescu , dennis shasha , eric simon , and cristian - augustin saita . declarative data cleaning : language , model , and algorithms . in _ proceedings of the international conference on very large databases ( vldb ) _ , pages 371380 , 2001 . luis gravano , panagiotis g. ipeirotis , h. v. jagadish , nick koudas , s. muthukrishnan , and divesh srivastava . approximate string joins in a database ( almost ) for free . in _ proceedings of the 27th international conference on very large data bases ( vldb ) _ , pages 491500 , san francisco , ca , usa , 2001 . morgan kaufmann publishers inc . luis gravano , panagiotis g. ipeirotis , nick koudas , and divesh srivastava . text joins for data cleansing and integration in an rdbms . in _ proceedings of the 19th international conference on data engineering ( icde ) _ , pages 729731 , march 2003 . dan gusfield . . cambridge university press , new york , ny , usa , 1997 . mauricio a. hernndez and salvatore j. stolfo . real - world data is dirty : data cleansing and the merge / purge problem . 2(1):937 , 1998 . m. a. jaro . advances in record linkage methodology as applied to matching the 1985 census of tampa . , pages 414420 , 1984 . nick koudas , amit marathe , and divesh srivastava . flexible string matching against large databases in practice . in _ proceedings of the international conference on very large databases ( vldb ) _ , pages 10781086 , august 2004 . nick koudas , amit marathe , and divesh srivastava . spider : flexible matching in databases . in _ sigmod 05 : proceedings of the 2005 acm sigmod international conference on management of data _ , pages 876878 , new york , ny , usa , 2005 . acm press . nick koudas , amit marathe , and divesh srivastava . using spider : an experience report . in _ sigmod 06 : proceedings of the 2006 acm sigmod international conference on management of data _ , page 719 , 2006 . nick koudas , sunita sarawagi , and divesh srivastava . record linkage : similarity measures and algorithms . in _ sigmod 06 : proceedings of the 2006 acm sigmod international conference on management of data _ , pages 802803 , 2006 . nick koudas and divesh srivastava . approximate joins : concepts and techniques . in _ proceedings of the international conference on very large databases ( vldb ) _ , page 1363 , 2005 . david r. h. miller , tim leek , and richard m. schwartz . a hidden markov model information retrieval system . in _ proceedings of the 22nd annual international acm sigir conference on research and development in information retrieval _ , pages 214221 , august 1999 . jay m. ponte and w. bruce croft . a language modeling approach to information retrieval . in _ proceedings of the 21st annual international acm sigir conference on research and development in information retrieval _ , pages 275281 , august 1998 . l.r . rabiner . a tutorial on hidden markov models and selected applications inspeech recognition . in _ proceedings of the ieee _ , volume 77 , pages 257286 , 1989 . stephen robertson . understanding inverse document frequency : on theoretical arguments . , 60(5):503520 , 2004 . stephen e. robertson , steve walker , micheline hancock - beaulieu , mike gatford , and a. payne . okapi at trec-4 . in _ trec _ , 1995 . gerard salton and chris buckley . term - weighting approaches in automatic text retrieval . , 24(5):513523 , 1988 . gerard salton and michael j. mcgill . . mcgraw - hill , inc . , new york , ny , usa , 1986 . sunita sarawagi and alok kirpal . efficient set joins on similarity predicates . in sigmod04 : proceedings of the acm sigmod international conference on management of data _ , pages 743754 , june 2004 . w. e. winkler . the state of record linkage and current research problems . technical report rr99/04 , us bureau of the census , 1999 . we assume the base relation ` base_table ` has an integer tuple i d attribute ` tid ` and a string valued attribute ` string ` . the following sql statements tokenize the base relation , creating ` base_tokens(tid , token ) ` . assuming that the query relation ` query_table ` has a single string valued attribute ` string ` , the same sql statements can be used for tokenization of the query string by removing ` tid ` from the statements . the size of the qgrams . + insert into integers(i ) values ( 1 ) , ( 2 ) , , ( max_str_size + @xmath110 ) + + insert into base_tokens(tid , token ) + select tid , substring(concat(substring($ $,1,@xmath11 - 1 ) , + upper ( replace(concat(string ) , , substring($ $,1,@xmath11 - 1 ) ) ) , + substring($ $,1,@xmath11 - 1 ) ) , integers.i , @xmath11 ) + from integers inner join base_table on + integers.i < = length ( replace(concat(string ) , , substring($ $,1,@xmath11 - 1 ) ) ) + ( @xmath11 - 1 ) + select & tid , substring(concat(string ) , 1 , locate( , concat(string ) ) - 1 ) + from & base_table + where & locate( , concat(string ) ) > 0 + + select & tid , substring(concat(string ) , n1.i+1 , n2.i - n1.i-1 ) + from & base_table , integers n1 , integers n2 + where & n1.i = locate( , concat(string ) , n1.i ) and n2.i = locate( , concat(string ) , n1.i + 1 ) + + select & tid , substring(concat(string ) , length(concat(string ) ) - locate( , reverse(concat(string)))+2 ) + from & base_table + where & locate( , concat(string ) ) > 0 + + select & tid , concat(string ) + from & base_table + where & locate( , concat(string ) ) = 0 + select & tid , token , + & substring(concat(substring($ $,1,@xmath11 - 1 ) , upper(token ) , substring($ $,1,@xmath11 - 1 ) ) , integers.i , @xmath11 ) + from & integers inner join base_tokens on integers.i < = length(token ) + ( @xmath11 - 1 ) + group by & tid , token , qgram + + select & t.tid , count ( * ) + from & base_tokens t + group by & t.tid + & + + select & t.tid , t.token , d.ddl + from & base_tokens t , base_ddl d + where & t.tid = d.tid + + + select & s1.tid , count()/(s1.ddl + s2.ddl - count ( ) ) + from & base_tokensddl s1 , query_tokens r2 , ( select count ( * ) as ddl from query_tokens t ) s2 + where & s1.token = r2.token + group by & s1.tid + + select & count ( * ) + from & base_table + & + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.token , log(s.size - count(t.tid ) + 0.5 ) - log(count(t.tid ) + 0.5 ) + from & base_tf t , base_size s + group by & t.token + & + + select & t.tid , t.token , i.midf + from & base_bmidf i , base_tf t + where & i.token = t.token + + + select & w1.tid , t2.tid , sum(w1.weight ) + from & base_weights w1 , query_tokens t2 + where & w1.token = t2.token + group by & t2.tid , w1.tid + + select & count ( * ) + from & base_table + & + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.token , log(s.size - count(t.tid ) + 0.5 ) - log(count(t.tid ) + 0.5 ) + from & base_tf t , base_size s + group by & t.token + & + + select & t.tid , t.token , i.midf + from & base_bmidf i , base_tokens t + where & i.token = t.token + & + + select & w.tid , sum(weight ) + from & base_weights w + group by & w.tid + & + + select & w.tid , w.token , d.ddl , w.weight + from & base_weights w , base_ddl d + where & w.tid = d.tid + + + select & s1.tid , sum(s1.weight)/(s1.ddl + s2.ddl - sum(s1.weight ) ) + from & base_tokensddl s1 , query_tokens r2 , + & ( select sum(t.weight ) as ddl + & from ( select t.token , i.idf as weight + & from base_idf i , query_tokens t + & where i.token = t.token ) t ) s2 + where & s1.token = r2.token + group & by s1.tid + + select & count ( * ) + from & base_table + & + + select & t.token , log(s.size ) - log(count(distinct t.tid ) ) + from & base_tokens t , base_size s + group by & t.token + & + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.tid , sqrt(sum(i.idfi.idft.tft.tf ) ) + from & base_idf i , base_tf t + where & i.token = t.token + group by & t.tid + & + + select & t.tid , t.token , i.idft.tf/l.len + from & base_idf i , base_tf t , base_length l + where & i.token = t.token and t.tid = l.tid + + select & r1w.tid , sum(r1w.weightr2w.weight ) + from & base_weights r1w , + & ( select t.token , qidf.idfqtf.tf/qlen.length as weight + & from ( select r.token , r.idf + & from query_tokens s , base_idf r + & where s.token = r.token + & group by s.token ) qidf , + & ( select t.token , count ( * ) as tf + & from query_tokens t + & group by t.tid , t.token ) qtf , + & ( select sqrt(sum(qidf.idfqidf.idfqtf.tfqtf.tf ) ) as length + & from ( select r.token , r.idf + & from query_tokens s , base_idf r + & where s.token = r.token + & group by s.token ) qidf , + & ( select t.token , count ( ) as tf + & from query_tokens t + & group by t.token ) qtf + & where i.token = t.token ) qlen + & where qidf.token = qtf.token ) r2w + where & r1w.token = r2w.token + group by & r1w.tid + + select & count ( * ) + from & base_table + & + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.token , log(s.size - count(t.tid ) + 0.5 ) - log(count(t.tid ) + 0.5 ) + from & base_tf t , base_size s + group by & t.token + & + + select & t.tid , sum(t.tf ) + from & base_tf t + group by & t.tid + & + + select & avg(len ) + from & base_bmbaselength + & + + select & t.tid , t.token , ( t.tf(@xmath43+1 ) ) / ( ( ( ( 1 - @xmath45)+(@xmath45l.dl / a.avgdl))@xmath43 ) + t.tf ) + from & base_bmbaselength l , base_bmbaseavglength a , base_tf t + where & l.tid = t.tid + & + + select & t.tid , t.token , t.mtfi.midf + from & base_bmbasemodtf t , base_bmidf i + where & t.token = i.token + + + select & b.tid , sum(b.weight * s.mtf ) + from & base_bmbaseweights b , + & ( select token , ( count())(@xmath44 + 1 ) / ( @xmath44+count ( * ) ) as mtf + & from query_tokens t + & group by t.token ) s + where & b.token = s.token + group by & b.tid + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.tid , count ( * ) + from & base_tokens t + group by & t.ti + & + + select & t.tid , t.token , t.tf/d.dl + from & base_tf t , base_dl d + where & t.tid=d.tid + & + + select & p.token , avg(p.pml ) + from & base_pml p + group by & p.token + & + + select & t.tid , t.token , p.pavgd.dl + from & base_tf t , base_pavg p , base_dl d + where & t.token = p.token and t.tid=d.tid + & + + select & t.tid , t.token , ( 1.0/(1.0+q.freq ) ) ( power(q.freq/(1.0+q.freq ) , t.tf ) ) + from & base_tf t , base_freq q + where & t.tid=q.tid and t.token = q.token + & + + select & count ( * ) + from & base_tokens + & + + select & t.token , count ( * ) / s.size + from & base_tokens t , base_tsize s + group by & t.token + & + select & t.tid , t.token , 1e0 * power(m.pml , 1.0-r.risk ) * power(a.pavg , r.risk ) , c.cfcs + from & base_tf t , base_risk r , base_pml m , base_pavg a , base_cfcs c , base_tsize s + where & t.tid=r.tid and t.token = r.token and t.tid=m.tid and + & t.token = m.token and t.token = a.token and t.token = c.token + & + + select & p.tid , sum(log(1.0-p.pm ) ) + from & base_pm p + group by & p.tid + + + select & b1.tid , b1.score + b2.sumcompm + from & ( select p1.tid , sum(log(p1.pm ) ) - sum(log(1.0-p1.pm ) ) - sum(log(p1.cfcs ) ) as score + & from base_pm p1 , query_tokens t2 + & where p1.token = t2.token + & group by p1.tid ) b1 , + & base_sumcompmbase b2 + where & b1.tid=b2.tid + + select & count ( * ) + from & base_table + & + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.tid , count ( * ) + from & base_tokens t + group by & t.tid + & + + select & t.tid , t.token , f.tf/d.dl + from & base_tokens t , base_tf f , base_dl d + where & f.tid=t.tid and t.token = f.token and t.tid=d.tid + & + + select & sum(t.dl ) + from & base_dl t + & + + select & t.token , sum(t.tf)/d.sdl + from & base_tf t , base_sumdl d + group by & t.token + & + + select & m.tid , m.token , log ( ( 1 + ( @xmath111m.pml ) / ( @xmath112p.ptge ) ) ) + from & base_ptge p , base_pml m + where & p.token = m.token + group by & tid , token + + + select & w1.tid , exp(sum(w1.weight ) ) + from & base_weights w1 , query_tokens t2 + where & w1.token = t2.token + + + select & count ( * ) + from & base_table + & + + select & t.token , log(s.size ) - log(count(distinct t.tid ) ) + from & base_tokens t , base_size s + group by & t.token + & + + select & avg(i.idf ) + from & base_idf i + & + + select & t.tid , t.token , count ( * ) + from & base_qgrams t + group by & t.tid , t.toke + & + + select & t.tid , t.token , t.qgram , s.size + from & base_qgrams t , base_tokensize s + where & t.tid = s.tid and t.token = s.token + select & maxsim.tid , r.string + & ( 1.0 - 1.0/@xmath11 ) + ( 1/sum(qidf.idf ) ) * sum(qidf.idf * ( ( ( 2.0/@xmath11 ) * maxsim.maxsim ) ) ) as score + from & ( select jac_sim.tid , jac_sim.token2 , max(sim ) as maxsim + & from ( select bsize.tid as tid , bsize.token as token1 , q.token as token2 , + & count()/(bsize.size + qsize.size - count ( ) ) as sim + & from base_qgramstokensize bsize , query_qgrams q , + & ( select t.token , count ( * ) as size + & from query_qgrams t + & group by t.token ) qsize + & where bsize.qgram = q.qgram and q.token = qsize.token + & group by bsize.tid , bsize.token , q.token ) jac_sim + & group by jac_sim.tid , jac_sim.token2 ) maxsim , + & ( select r.token , r.idf + & from query_tokens s , base_idf r + & where s.token = r.token + & group by s.token + & union + & select s.token , a.idfavg as idf + & from query_tokens s , base_idfavg a + & where s.token not in ( select i.token from base_idf i ) + & group by s.token ) qidf , + & base_table r + where & tm.token2 = i.token and r.tid = maxsim.tid + group by & tm.tid + having & score > = @xmath4 + + select & count ( * ) + from & base_table + & + + select & t.token , log(s.size ) - log(count(distinct t.tid ) ) + from & base_tokens t , base_size s + group by & t.token + & + + select & avg(i.idf ) + from & base_idf i + & + + select & n.i-1 , round(rand()maxint ) + from & integers n + limit & hash_size + & + + select & f.fid , q.qgram , mod(conv(hex ( q.qgram ) , 16 , 10 ) maxint , f.func ) + from & base_hashfunc f , ( select distinct qgram from base_qgrams ) q + & + + select & q.tid , q.token , h.fid , min(h.value ) + from & base_qgrams q , base_hashvalue h + where & q.qgram = h.qgram + group by & q.tid , q.token , h.fid + select & maxsim.tid , r.string , + & ( 1.0 - 1.0/@xmath11 ) + ( 1/sum(i.idf ) ) * sum(i.idf * ( ( ( 2.0/@xmath11 ) * maxsim.maxsim ) ) ) as score + from & ( select mh_sim.tid , mh_sim.token2 , max(sim ) as maxsim + & from ( select bmhsig.tid as tid , bmhsig.token as token1 , + & qmhsig.token as token2 , count()/@xmath115 as sim + & from base_minhashsignature bmhsig , + & ( select q.token , h.fid , min(h.value ) as value + & from query_qgrams q , base_hashvalue h + & where q.qgram = h.qgram + & group by q.token , h.fid ) qmhsig + & where bmhsig.fid = qmhsig.fid and bmhsig.value = qmhsig.value + & group by bmhsig.tid , bmhsig.token , qmhsig.token ) mh_sim + & group by mh_sim.tid , mh_sim.token2 ) maxsim , + & ( select r.token , r.idf + & from query_tokens q , base_idf r + & where q.token = r.token + & group by q.token + & union + & select q.token , a.idfavg as idf + & from query_tokens q , base_idfavg a + & where q.token not in ( select i.token from base_idf i ) + & group by q.token ) qidf , + & base_table r + where & maxsim.token2 = i.token and r.tid = maxsim.tid + group by & maxsim.tid + having & score > = @xmath4 + + select & count ( ) + from & base_table + & + + select & t.token , log(s.size ) - log(count(distinct t.tid ) ) + from & base_tokens t , base_size s + group by & t.token + & + + select & t.tid , t.token , count ( * ) + from & base_tokens t + group by & t.tid , t.token + & + + select & t.tid , sqrt(sum(i.idfi.idft.tft.tf ) ) + from & base_idf i , base_tf t + where & i.token = t.token + group by & t.tid + & + + select & t.tid , t.token , i.idft.tf/l.len + from & base_idf i , base_tf t , base_length l + where & i.token = t.token and t.tid = l.tid + + + select & maxtoken.tid , sum(wb.weight * wq.weight * maxtoken.maxsim ) + from & base_weights wb , + & ( select jaro_sim.tid , jaro_sim.token1 , jaro_sim.token2 , maxsim.maxsim + & from ( select jaro_sim.tid , jaro_sim.token2 , max(sim ) as maxsim + & from ( select r1.tid as tid , r1.token as token1 , + & r2.token as token2 , jarowinkler(r1.token , r2.token ) as sim + & from base_tokens r1 , query_tokens r2 + & where jarowinkler(r1.token , r2.token ) > = @xmath4 ) jaro_sim + & group by jaro_sim.tid , jaro_sim.token2 ) maxsim , + & ( select r1.tid as tid , r1.token as token1 , + & r2.token as token2 , jarowinkler(r1.token , r2.token ) as sim + & from base_tokens r1 , query_tokens r2 + & where jarowinkler(r1.token , r2.token ) > = @xmath4 ) jaro_sim + & where jaro_sim.tid = maxsim.tid and jaro_sim.token2 = maxsim.token2 + & and maxsim.maxsim = jaro_sim.sim ) maxtoken , + & ( select qtf.token , qidf.idfqtf.tf/qlen.length as weight + & from ( select r.token , r.idf + & from query_tokens s , base_idf r + & where s.token = r.token + & group by s.token ) qidf , + & ( select t.token , count ( ) as tf + & from query_tokens t + & group by t.token ) qtf , + & ( select qtf.tid , sqrt(sum(qidf.idfqidf.idfqtf.tfqtf.tf ) ) as length + & from ( select r.token , r.idf + & from query_tokens s , base_idf r + & where s.token = r.token + & group by s.token ) qidf , + & ( select t.token , count ( ) as tf + & from query_tokens t + & group by t.token ) qtf + & where qidf.token = qtf.token ) qlen + & where qidf.token = t.token ) wq , + where & maxtoken.token2 = wq.token and maxtoken.tid = wb.tid and maxtoken.token1 = wb.token + group by & maxtoken.tid +	declarative data quality has been an active research topic . the fundamental principle behind a declarative approach to data quality is the use of declarative statements to realize data quality primitives on top of any relational data source . a primary advantage of such an approach is the ease of use and integration with existing applications . over the last couple of years several similarity predicates have been proposed for common quality primitives ( approximate selections , joins , etc . ) and have been fully expressed using declarative sql statements . in this thesis , new similarity predicates are proposed along with their declarative realization , based on notions of probabilistic information retrieval . then , full declarative specifications of previously proposed similarity predicates in the literature are presented , grouped into classes according to their primary characteristics . finally , a thorough performance and accuracy study comparing a large number of similarity predicates for data cleaning operations is performed .
, coupling strength @xmath0 ) . @xmath1 is the nel temperature , while the temperature below which superconductivity ( a pseudogap ) is observed is denoted by @xmath2 ( @xmath3 ) . [ from ref . @xcite . ] [ phase - diagram - exp ] ] after intensive investigations over more than two decades , it has become clear that cuprate superconductors are among the most complicated systems studied in condensed matter physics @xcite . the complications arise mainly from that the parent compounds of cuprate superconductors are a form of non - conductor called a mott insulator with an antiferromagnetic ( af ) long - range order ( aflro ) @xcite , where a single common feature in the layered crystal structure is the presence of one to several cuo@xmath4 planes in the unit cell @xcite . inelastic neutron scattering ( ins ) experiments show that the low - energy spin excitations in these parent compounds are well described by an af heisenberg model @xcite with the magnetic exchange coupling constant @xmath5 ev . when these cuo@xmath4 planes are doped with charge carriers , the aflro phase subsides and superconductivity emerges leaving the af short - range order ( afsro ) correlation still intact @xcite . although there are hundreds of cuprate superconducting ( sc ) compounds , they all fit into a universal phase diagram @xcite as schematically illustrated in fig . [ phase - diagram - exp ] , where the physical properties mainly depend on the extent of doping , and the regimes have been classified into the underdoped , optimally doped , and overdoped , respectively . after aflro is destroyed rapidly by doping , there are three apparent regions of the phase diagram : ( a ) a d - wave sc phase , where the maximal sc transition temperature @xmath2 occurs around the optimal doping , and then decreases in both the underdoped and overdoped regimes @xcite ; ( b ) a normal - state pseudogap metallic phase , where an energy gap called the normal - state pseudogap @xmath6 exists @xcite above @xmath2 but below the normal - state pseudogap crossover temperature @xmath3 . however , in contrast to the domelike shape of the doping dependence of @xmath2 , @xmath3 is much larger than @xmath2 in the underdoped and optimally doped regimes @xcite , and then monotonically decreases upon the increase of doping . in particular , measurements taken by using a wide variety of techniques demonstrate that the normal - state pseudogap is present in both the spin and charge channels @xcite ; ( c ) a _ normal - metal _ phase with largely transport properties . in the doped regime , charge carriers couple to spin excitations @xcite . the combined ins and resonant inelastic x - ray scattering ( rixs ) experimental data have identified spin excitations with high intensity over a large part of moment space , and shown that spin excitations exist across the entire range of the sc dome @xcite . however , the charge - carrier doping causes substantial changes to the low - energy spin excitation spectrum @xcite , while it has a more modest effect on the high - energy spin excitations @xcite . in particular , rixs experiments @xcite show that the high - energy spin excitations persist well into the overdoped regime and bear a striking resemblance to those found in the parent compounds , indicating that a local - moment picture accounts for the observed spin excitations at elevated energies even up to the overdoped regime @xcite . experimentally , a large body of data available from a wide variety of measurement techniques have provided rather detailed information on cuprate superconductors , where some essential agreements have emerged . we refer the readers to the more detailed summaries of experimental results available in the literatures @xcite . superconductivity , the dissipationless flow of electrical current , is a striking manifestation of a subtle form of quantum rigidity on the macroscopic scale @xcite , where a central question is what mechanism causes the loss of electrical resistance below @xmath2 ? it is commonly believed that the existence of electron cooper pairs is the hallmark of superconductivity @xcite , since these electron cooper pairs behave as effective bosons , and can form something analogous to a bose condensate that flows without resistance . this follows from a fact that although electrons repel each other because of the coulomb interaction , at low energies there can be an effective attraction that originates by the exchange of bosons @xcite . in conventional superconductors , as explained by the bardeen - cooper - schrieffer ( bcs ) theory @xcite , these exchanged bosons are phonons that act like a bosonic _ glue _ to hold the electron pairs together , and then these electron pairs condense into a coherent macroscopic quantum state that is insensitive to impurities and imperfections and hence conducts electricity without resistance @xcite . the excitation in the sc - state has an energy gap @xmath7 , which determines both the quasiparticle energy spectrum and the energy of the condensate @xcite . in this conventional electron - phonon sc mechanism @xcite , the resulting wave function for the pairs turns out to be peaked at zero separation of the electrons , which leads to that the sc - state has an s - wave symmetry @xcite . as a consequence , the pairs in conventional superconductors are always related to an increase in kinetic energy which is overcompensated by the lowering of potential energy @xcite . at the temperature above @xmath2 , the electron is in the standard landau fermi - liquid state , which is generally referred to as a _ normal - state _ , where the density of states near the fermi level is smooth and generally treated as featureless @xcite . as in conventional superconductors , superconductivity in cuprate superconductors results when charge carriers pair up into charge - carrier pairs , which is supported by many experimental evidences , including the factor of @xmath8 occurring in the flux quantum and in the josephson effect , as well as the electrodynamic and thermodynamic properties @xcite . however , the normal - state of cuprate superconductors in the pseudogap phase is not normal at all , since the normal - state of cuprate superconductors in the pseudogap phase exhibits a number of the anomalous properties @xcite in the sense that they do not fit in with the standard landau fermi - liquid theory . superconductivity is an instability of the normal - state . however , one of the most striking dilemmas is that the sc coherence of quasiparticle peaks in cuprate superconductors is described by a standard bcs formalism , although the normal - state is undoubtedly not the standard landau fermi - liquid on which the conventional bcs electron - phonon sc mechanism is based . angle - resolved photoemission spectroscopy ( arpes ) experiments reveal sharp spectral peaks in the excitation spectrum @xcite , indicating the presence of quasiparticle - like states , which is also consistent with the long lifetime of electronic state as it has been determined by the conductivity measurements @xcite . moreover , arpes experiments also observe the bogoliubov - type dispersion of the sc - state @xcite predicted by the standard bcs formalism . however , as a natural consequence of the unconventional sc mechanism that is responsible for the high @xmath2 , the charge - carrier pairs in cuprate superconductors have a dominant d - wave symmetry @xcite . this d - wave sc - state also implies that there is a strongly momentum - dependent attraction between charge carriers without phonons @xcite . after over more than two decades of the painstaking effort , people are still debating the very mechanism of superconductivity in cuprate superconductors , where the crucial issues in cuprate superconductors are ( a ) what is the nature of the _ glue _ binding charge carriers into charge - carrier pairs , such that they can travel macroscopic distances without resistance ? ( b ) whether the pseudogap has a competitive or collaborative role in engendering superconductivity ? very soon after the discovery of superconductivity in cuprate superconductors @xcite , anderson @xcite proposed that in the parent compounds of cuprate superconductors , the spins form a superposition of singlets . this spin liquid of singlets is so - called the resonating valence bond ( rvb ) state . upon the charge - carrier doping , these rvb singlets would become charged , resulting in a sc - state . the rvb state is fundamentally different from the conventional nel state in which the doped charge carrier can move freely among the rvb spin liquid and then a better compromise between the charge - carrier kinetic energy and spin exchange energy may be achieved . since then many elaborations of this idea followed @xcite . in particular , it was realized that essential aspects of the rvb concept can be formulated within the charge - spin separation ( css ) slave - particle approach @xcite , since the essential physics of cuprate superconductors is dominated by the short - range repulsive interaction which remains relevant and causes css @xcite . this css slave - particle approach also led to the development of the gauge theory for cuprate superconductors @xcite . however , in the framework of the original rvb theory @xcite , the af exchange coupling @xmath9 attracts electrons of opposite spins to be on neighboring sites . this is the result of states of very high energy with a spin gap , and the corresponding interaction has only high - energy dynamics @xcite . the normal - state pseudogap is identified as the spin gap in the rvb state with an energy scale set by @xmath9 , and therefore is associated with the breaking of the rvb singlets @xcite . in this case , anderson @xcite suggested that the sc - state in cuprate superconductors is determined by the need to reduce the frustrated kinetic energy of the system , where the strong frustration of the kinetic energy in the normal - state is partially relieved upon entering the sc - state , indicating that the kinetic energy causes superconductivity @xcite . on the other hand , it has been argued phenomenologically that superconductivity in cuprate superconductors could arise from a lowering of the kinetic energy rather than the potential energy @xcite . in this scenario , superconductors would exhibit qualitatively new features in their optical properties a violation of the low - energy optical sum rule , and a change in the high - energy optical absorption when the system becomes sc @xcite . later , the high - precision optical measurements on cuprate superconductors in the near - infrared and visible region indicate small changes in the spectral weight associated with the onset of superconductivity @xcite , therefore supporting this argument @xcite that changes in the kinetic energy are indeed occurring . in particular , the similar experimental results have by now been obtained by differently experimental groups @xcite . more importantly , the recent experimental results @xcite from the arpes measurements on cuprate superconductors indicate that @xmath2 is correlated with the charge - carrier kinetic energy , which supports the notion of the kinetic - energy driven superconductivity . motivated by these experimental results @xcite , several calculations based on the strongly correlated models have been done to show that superconductivity may be driven by a lowering of the kinetic energy upon the formation of the sc - state @xcite . by constructing an effective hamiltonian for spin polarons forming in weakly doped antiferromagnets , it has been demonstrated that the driving mechanism which gives rise to superconductivity in such system is the reduction of the kinetic energy @xcite . moreover , a numerical study of the two - dimensional hubbard model within the dynamical cluster approximation has shown the lowering of the kinetic energy below @xmath2 for different doping levels @xcite . these theoretical calculation @xcite shows that the paired charge carriers in the afsro background can be more mobile than the single charge carriers , and this can overcome the normal increase in the kinetic energy upon the pair formation @xcite . superconductivity in cuprate superconductors is something entirely new , a manifestation of the strong electron correlation , or mottness @xcite . in the early days of superconductivity , we @xcite have developed a fermion - spin theory to confront the strong electron correlation , where the constrained electron operator is decoupled as a product of a charge carrier and a localized spin , with the charge carrier represented the charge degree of freedom of the electron together with some effects of spin configuration rearrangements due to the presence of the doped charge carrier itself , while the spin operator represented the spin degree of freedom of the electron , and then the strong electron correlation can be treated properly in actual calculations . in particular , these charge carriers and spin are gauge invariant , and in this sense , the collective modes for these charge carriers and spins are real and can be interpreted as the physical excitations of the system . in this fermion - spin theory , the basic low - energy excitations are charge - carrier quasiparticles , the spin excitations , and the electron quasiparticles . in this case , the charge transport is mainly governed by the scattering of charge carriers due to spin fluctuations , and the scattering of spins due to charge - carrier fluctuations dominates the spin dynamics , while as a result of the charge - spin recombination , the electron quasiparticles are responsible for the electronic properties . within the framework of the fermion - spin theory @xcite , we have established a kinetic - energy driven sc mechanism @xcite , where charge carriers are held together in d - wave pairs at low temperatures by the attractive interaction in the particle - particle channel that originates directly from the kinetic energy by the exchange of spin excitations in the higher powers of the doping concentration , and then these charge - carrier pairs ( then electron cooper pairs ) condense to the d - wave sc - state . although the physical properties of cuprate superconductors in the normal - state are fundamentally different from these in the standard landau fermi - liquid state , the kinetic - energy driven sc - state still is conventional bcs - like with the d - wave symmetry , and then the obtained formalism for the charge - carrier pairing can be used to compute @xmath2 and the related sc coherence of the low - energy excitations in cuprate superconductors on the first - principles basis much as can be done for conventional superconductors . moreover , this kinetic - energy driven sc - state is controlled by both the charge - carrier pair gap and quasiparticle coherence , which leads to that @xmath2 takes a domelike shape with the underdoped and overdoped regimes on each side of the optimal doping @xmath10 , where @xmath2 reaches its maximum . on the other hand , the same charge - carrier interaction mediated by spin excitations that induces the sc - state in the particle - particle channel also generates the normal - state pseudogap state in the particle - hole channel @xcite . as a consequence , the sc gap and normal - state pseudogap coexist but compete in the whole sc dome however , the normal - state pseudogap crossover temperature @xmath3 is much larger than @xmath2 in the underdoped and optimally doped regimes , and then monotonically decreases upon the increase of doping , eventually disappearing together with superconductivity at the end of the sc dome . this kinetic - energy driven sc mechanism therefore provides a natural explanation of both the origin of the normal - state pseudogap state and pairing mechanism for superconductivity . it is beyond the scope of this article to provide an overview of various theories of superconductivity in cuprate superconductors that have been put forward in the literatures , and some earlier reviews and different perspectives appear in refs . @xcite . in this article , we only attempt to review comprehensively the general framework of the kinetic - energy driven sc mechanism in the context of our work @xcite and to summarize several calculated results of physical quantities obtained based on the kinetic - energy driven sc mechanism . the number of topics for this review article is listed as follows . in section [ cssfst ] , we @xcite give an overview of the fermion - spin theory , and show that within the _ decoupling _ scheme , the fermion - spin representation is a natural representation of the constrained electron defined in a restricted hilbert space without double electron occupancy . the kinetic - energy driven sc mechanism @xcite is introduced in section [ kedm ] . in the rest of sections , we show how this kinetic - energy driven sc mechanism yields many results that are in broad agreement with various key experimental facts observed on cuprate superconductors . superconductors are not only perfect conductors , but also exhibit the so - called meissner effect , where they expel magnetic fields . in section [ meissner - effect ] , we consider main features of the doping dependence of the electromagnetic response observed on cuprate superconductors by using the muon - spin - rotation measurement technique , and show that in analogy to the domelike shape of the doping dependence of @xmath2 , the maximal superfluid density @xmath11 occurs around the critical doping @xmath12 , and then decreases in both lower doped and higher doped regimes . in section [ spin - response ] , we turn to the comparison of the calculated result of the dynamical spin response with rixs - ins experimental data . it is shown that the low - energy spin excitations in the sc - state have an hour - glass - shaped dispersion , with commensurate resonance that appears in the sc - state _ only _ , while the low - energy incommensurate ( ic ) spin fluctuations can persist into the normal - state . the high - energy spin excitations in the sc - state on the other hand retain roughly constant energy as a function of doping , with spectral weights and dispersion relations comparable to those found in the parent compounds . a brief description of the interplay between superconductivity and normal - state pseudogap state is given in section [ sc - pseudogap ] , where we @xcite identify the normal - state pseudogap as being a region of the self - energy effect in the particle - hole channel in which the normal - state pseudogap suppresses the spectral weight of the low - energy excitation spectrum . this normal - state pseudogap disappears at @xmath3 , and then system crossovers to the normal - metal phase with largely transport properties at the temperatures @xmath13 . in section [ charge - transport ] , we discuss the effect of the normal - state pseudogap on the infrared response of cuprate superconductors , and show that in the underdoped and optimally doped regimes , the transfer of the part of the low - energy spectral weight of the conductivity spectrum to the higher energy region to form a midinfrared band is intrinsically associated with the emergence of the normal - state pseudogap . finally , the article concludes with the suggestions , in section [ conclusion ] , for future work . in cuprate superconductors , the single common feature in the layered crystal structure is the presence of one to several cuo@xmath4 planes in the unit cell @xcite , and it seems evident that the nonconventional behaviors of cuprate superconductors are dominated by the cuo@xmath4 plane . in this case , as originally emphasized by anderson @xcite , the essential physics of the doped cuo@xmath4 plane is contained in the one - band large-@xmath14 hubbard model @xcite on a square lattice , @xmath15 where the summation is over all sites @xmath16 , and the hopping integrals @xmath17 connect sites @xmath16 and @xmath18 . we will restrict to our attention to the nearest ( @xmath19 ) and next nearest ( @xmath20 ) neighbor hopping . @xmath21 and @xmath22 are electron operators that respectively create and annihilate electrons with spin @xmath23 , @xmath24 , and @xmath25 is the chemical potential . this large-@xmath14 hubbard model ( [ hubbard ] ) indicates that the interactions in cuprate superconductors are dominated by the on - site mott - hubbard term @xmath14 , which is very large as compared with the electron hopping integrals @xmath19 and @xmath26 , i.e. , @xmath27 , @xmath26 , and therefore leads to that electrons become strongly correlated to avoid double occupancy . in this case , the on - site mott - hubbard term must be dealt properly before bothering with relatively minor terms @xcite . it has been shown @xcite that the correct way to deal with this large-@xmath14 term in eq . ( [ hubbard ] ) is to renormalize it by means of a canonical transformation @xmath28 , which eliminates large-@xmath14 term from the block which contains no doubly occupied states , and which presumably contains all the low - energy eigenstates and thus the ground state , and then the transformed hamiltonian can be obtained as , @xmath29 with the nearest - neighbors @xmath30 , the next nearest - neighbors @xmath31 , the magnetic exchange coupling constant @xmath32 , the spin operators @xmath33 . the kinetic - energy term in eq . ( [ tjmodel ] ) describes mobile charge carriers in the af background , while the heisenberg term in eq . ( [ tjmodel ] ) describes af coupling between localized spins . in particular , the nearest - neighbor hopping integral @xmath19 in the kinetic - energy term is much larger than the magnetic exchange coupling constant @xmath9 in the heisenberg term , and therefore the spin configuration is strongly rearranged due to the effect of the charge - carrier hopping @xmath19 on the spins , which leads to strong coupling between the charge and spin degrees of freedom of the electron . this transformed hamiltonian ( [ tjmodel ] ) is so - called @xmath19-@xmath9 model acting on a restricted hilbert space without double electron occupancy , where there are three physical states _ only _ , latexmath:[\ ] ] where the spectral functions @xmath560 and @xmath561 are obtained in terms of the charge - carrier diagonal and off - diagonal green s functions in eq . ( [ hgf-1 ] ) as @xmath562 and @xmath563 , respectively . with @xmath248 for @xmath246 and @xmath247 . inset : the corresponding experimental data of the underdoped bi@xmath4sr@xmath4cacu@xmath4o@xmath257 taken from ref . [ from ref . @xcite . ] [ sc - conductivity ] ] in fig . [ sc - conductivity ] , we @xcite show the calculated result of the conductivity ( [ conductivity ] ) in the sc - state as a function of energy at @xmath250 for @xmath246 and @xmath247 with @xmath248 in comparison with the corresponding experimental result @xcite of the underdoped bi@xmath4sr@xmath4cacu@xmath4o@xmath257 ( inset ) . the result in fig . [ sc - conductivity ] shows clearly that the two - component feature of the conductivity spectrum in the sc - state is the same as in the normal - state case . in particular , in comparison with the result of the conductivity in the normal - state in subsection [ normal - conductivity ] , it is found that the spectral weight of the low - energy component of the conductivity in the sc - state is further suppressed by the sc gap , however , there is no depletion of the spectral weight of the higher energy midinfrared band of the conductivity in the sc - state , which is consistent with the experimental observation on cuprate superconductors @xcite . in this case , although there is a coexistence of the sc gap and normal - state pseudogap , the onset of the region to which the spectral weight is transferred is also close to the normal - state pseudogap @xmath6 , then in analogy to the evolution of the conductivity spectrum with doping in the normal - state , the positions of the gap in the conductivity spectrum and midinfrared peak gradually shift to the lower energies with increasing doping @xcite . all the calculated results @xcite of the conductivity spectrum in the sc - state are also qualitatively consistent with the corresponding experimental data of cuprate superconductors in the sc - state @xcite . in the sc - state , there are two parts of the charge - carrier quasiparticle contribution to the redistribution of the spectral weight in the conductivity spectrum in the sc - state : the contribution from the first term of the right - hand side in eq . ( [ conductivity ] ) comes from the spectral function obtained in terms of the charge - carrier diagonal green s function ( [ hdgf-1 ] ) , and therefore is closely associated with the normal - state pseudogap @xmath6 in the particle - hole channel , while the additional contribution from the second term of the right - hand side in eq . ( [ conductivity ] ) originates from the spectral function obtained in terms of the charge - carrier off - diagonal green s function ( [ hodgf-1 ] ) , and is closely related to the charge - carrier pair gap @xmath242 in the particle - particle channel . however , since @xmath564 in the underdoped and optimally doped regimes as we have mentioned in section [ sc - pseudogap ] , the charge - carrier pair gap only suppresses the spectral weight of the low - energy component , while the normal - state pseudogap related shift of the spectral weight from the low - energy to the higher energy midinfrared band in the sc - state conductivity spectrum becomes arrested . the effect of the normal - state pseudogap on the infrared response in cuprate superconductors has been also discussed based on the preformed pair theory @xcite and the phenomenological theory of the normal - state pseudogap state @xcite , and the results of the unusual two - component conductivity spectrum are qualitatively consistent with the above obtained result based on the kinetic - energy driven sc mechanism by considering the interplay between the sc gap and normal - state pseudogap . in particular , their results @xcite also indicate that in the underdoped and optimally doped regimes , the transfer of the part of the low - energy spectral weight of the conductivity spectrum to the higher energy region to form a midinfrared band is intrinsically associated with the presence of the normal - state pseudogap . in a standard landau fermi - liquid , the shape of the conductivity spectrum @xmath544 is normally well accounted for by the low - energy drude formula that describes the free charge carrier contribution to @xmath543 , and then when the temperatures @xmath565 , the spectral weight of the condensate in the sc - state comes from low energies @xcite . however , in cuprate superconductors , the part of the low - energy spectral weight in the conductivity spectrum in the underdoped and optimally doped regimes is transferred to the higher energy region to form the unusual midinfrared band , and then the width of the low - energy band is narrowing , while the onset of the region to which the spectral weight is transferred is close to the normal - state pseudogap @xmath6 . moreover , since the unusual midinfrared band is taken from the low - energy band , so that both the low - energy non - drude peak ( the conductivity decays as @xmath529 at low energies ) and unusual midinfrared band describe the actual charge - carrier density . in the framework of the kinetic - energy driven sc mechanism , the normal - state pseudogap state is the result of the strong electron correlation , and therefore the transfer of the part of the low - energy spectral weight of the conductivity spectrum in the underdoped and optimally doped regimes to the higher energy region to form the unusual midinfrared band is a natural consequence of the strongly correlated nature in cuprate superconductors . in particular , this strong electron correlation which induces a shift of the spectral weight from the low - energy to the higher energy midinfrared band in the conductivity spectrum , has been confirmed by the early numerical simulations based on the @xmath19-@xmath9 model in the normal - state @xcite and in the sc - state @xcite . in the normal - state , the @xmath566 perfect decay of the conductivity at low energies in the optimally doped regime is closely related with the linear temperature resistivity , since it reflects an anomalous frequency dependent scattering rate proportional to @xmath214 instead of @xmath567 as would be expected in the standard landau fermi - liquid . this linear temperature resistivity is one of the characteristically anomalous properties of cuprate superconductors in the normal - state , and has been also phenomenological discussed within the marginal fermi - liquid theory @xcite . in particular , based on the slave - boson gauge theory , it has been shown within the @xmath19-@xmath9 model that above the bose - einstein temperature , the boson inverse lifetime due to scattering by the gauge field is of order @xmath335 , which suppresses the condensation temperature and leads to a linear @xmath335 resistivity @xcite . however , in the sc - state , the large normal - state pseudogap in the underdoped and optimally doped regimes heavily reduces the fraction of the charge carriers that condense in the sc - state @xcite . the calculated result of the conductivity spectrum summarized in this section shows very clearly that if the effect of the normal - state pseudogap is taken into account in the framework of the kinetic - energy driven sc mechanism , the conductivity of the @xmath19-@xmath9 model calculated based on the linear response approach per se can correctly reproduce the main features found in infrared response measurements on cuprate superconductor in both the normal- and sc - states . the conductivity spectrum in the underdoped and optimally doped regimes contains the low - energy non - drude peak and unusual midinfrared band . however , the position of the midinfrared band shifts towards to the low - energy non - drude peak with increasing doping . in particular , the low - energy non - drude peak incorporates with the midinfrared band in the heavily overdoped regime , and then the low - energy drude behavior recovers . the qualitative reproduction of the main features of infrared response measurements on cuprate superconductors also shows that the transfer of the part of the low - energy spectral weight in the conductivity spectrum in the underdoped and optimally doped regimes to the higher energy region to form the unusual midinfrared band can be attributed to the effect of the normal - state pseudogap on the infrared response in cuprate superconductors . in this article , we have given a brief review of the kinetic - energy driven sc mechanism , where the main conclusions are summarized as : \(a ) in the fermion - spin theory ( [ css ] ) , the constrained electron is decoupled as a product of a charge carrier and a localized spin , and the charge carrier represents the charge degree of freedom of the constrained electron together with some effects of the spin configuration rearrangements due to the presence of the doped charge carrier itself , while the spin operator represents the spin degree of freedom of the constrained electron . in the _ decoupling scheme _ , this fermion - spin representation ( [ css ] ) is a natural representation of the constrained electron defined in a restricted hilbert space without double electron occupancy . the main advantage of the fermion - spin theory ( [ css ] ) is that the electron local constraint for single occupancy is satisfied in actual calculations . in particular , the charge carrier _ or _ spin _ itself _ is @xmath54 gauge invariant , and in this sense , the collective modes for the charge carrier and spin are real and can be interpreted as the physical excitations of cuprate superconductors . although both charge carriers and spins contribute to the charge and spin dynamics , the charge - carrier relaxation time is responsible to the charge transport , and the spin relaxation time is responsible to the dynamical spin response , while as a result of the charge - spin recombination , the the electronic properties are dominated by electron quasiparticles . this is an efficient calculation scheme which can provide very good results even at the mf level . \(b ) in the framework of the kinetic - energy driven sc mechanism developed based on the fermion - spin theory ( [ css ] ) , the charge - carrier pairing interaction originates directly from the kinetic energy of the @xmath19-@xmath9 model ( [ csstjmodel ] ) by the exchange of spin excitations in the higher powers of the doping concentration , and then these charge - carrier pairs ( then the electron cooper pairs ) condense to the d - wave sc - state , where the spin excitation in cuprate superconductors plays a similar role to that of the phonon in conventional superconductors . although the physical properties of cuprate superconductors in the normal - state are fundamentally different from these in the standard landau fermi - liquid state , the kinetic - energy driven sc - state still is conventional bcs - like with the d - wave symmetry , and then the obtained formalism for the charge - carrier pairing can be used to compute @xmath2 and the related sc coherence of the low - energy excitations in cuprate superconductors on the first - principles basis much as can be done for conventional superconductors . moreover , the kinetic - energy driven charge - carrier pair state is controlled by both the charge - carrier pair gap and quasiparticle coherence , which leads to that the maximal @xmath2 occurs around the optimal doping , and then decreases in both the underdoped and overdoped regimes . this kinetic - energy driven sc mechanism also indicates that the strong electron correlation favors superconductivity , since the main ingredient is identified into a charge - carrier pairing mechanism not from the external degree of freedom such as the phonon , but rather solely from the internal spin degree of freedom of the constrained electron . \(c ) the same charge - carrier interaction arising through the exchange of spin excitations that induces the d - wave sc - state in the particle - particle channel also generates the normal - state pseudogap state in the particle - hole channel , therefore there is a coexistence of the sc gap and normal - state pseudogap in the whole sc dome . consequently , this normal - state pseudogap is identified as being a region of the self - energy effect in the particle - hole channel in which the normal - state pseudogap suppresses the spectral weight of the low - energy excitation spectrum . this normal - state pseudogap vanishes at the normal - state pseudogap crossover temperature @xmath3 , with @xmath3 that is much larger than @xmath2 in the underdoped and optimally doped regimes , and monotonically decreases upon the increase of doping , eventually disappearing together with @xmath2 at the end of the sc dome . in particular , the normal - state pseudogap is directly related to the quasiparticle coherence , and therefore antagonizes superconductivity . moreover , both the normal - state pseudogap and charge - carrier pair gap are dominated by one energy scale , and they are the result of the strong electron correlation in cuprate superconductors , while the domelike shape of the doping dependence of @xmath2 , the monotonic decrease of @xmath3 with doping , and relatively anomalous normal - state properties are a natural consequence of the mott physics in which double occupancy is suppressed by strongly coulombic repulsion . the theory also indicates that the kinetic - energy driven sc mechanism provides a natural explanation of both the origin of the normal - state pseudogap state and the charge - carrier pairing mechanism for superconductivity . \(d ) within the framework of the kinetic - energy driven sc mechanism , a number of typical properties of cuprate superconductors have been studied . in this review article , the selected results are summarized , including the doping dependence of the electromagnetic response , the dynamical spin response from low - energy to high - energy , and the charge transport , and are qualitatively comparable to the corresponding experimental results observed in cuprate superconductors . furthermore , this kinetic - energy driven bcs - type formalism gives an explanation of the raman scattering spectra @xcite obtained in terms of the electronic raman response measurement technique @xcite . it also gives a consistent description of the thermodynamic properties @xcite observed from heat capacity measurements @xcite . in particular , it has been used to successfully describe a number of the sc - state properties in the presence of impurities @xcite , including the microwave conductivity @xcite , the scanning tunneling microscopic measurements of the coherent bogoliubov quasiparticle dispersion , and the related extinction of bogoliubov quasiparticle scattering interference at low temperatures @xcite . establishing these agreements between the calculated results obtained based on the kinetic - energy driven sc mechanism and the corresponding experimental data observed from a wide variety of measurement techniques are important to confirm the nature of the sc phase of cuprate superconductors to be the kinetic - energy driven d - wave sc - state . in this review article , we have restricted our attention to the hole - doped cuprate superconductors . however , superconductivity in cuprates also emerges when electrons are doped into mott insulators @xcite . both the hole - doped and electron - doped cuprate superconductors have the layered structure of the square lattice of the cuo@xmath4 plane separated by insulating layers @xcite . although the significantly different behaviors of the hole - doped and electron - doped cuprate superconductors are observed due to the electron - hole asymmetry , the symmetry of the sc order parameter is common in both case @xcite , manifesting that two systems have similar underlying sc mechanism . in particular , the strong electron correlation is common for both the hole - doped and electron - doped cuprate superconductors , and then it is possible that superconductivity in the electron - doped cuprate superconductors is also driven by the kinetic energy as in the hole - doped case . in this case , the charge asymmetry @xcite in superconductivity of the electron - doped and hole - doped cuprate superconductors and the electronic raman response @xcite in the electron - doped cuprate superconductors have been discussed based on the kinetic - energy driven sc mechanism , and the calculated results are in qualitative agreement with the experimental data observed on the electron - doped cuprate superconductors . besides the square lattice cuprate superconductors , some cuprate materials @xcite , such as sr@xmath568cu@xmath569o@xmath570 , do not contain cuo@xmath4 planes common to cuprate superconductors but consist of two - leg cu@xmath4o@xmath251 ladders and edge - sharing cuo@xmath4 chains . in particular , the doped two - leg ladder cuprates are a system in which the sc state is realized by applying a high pressure in the highly charge - carrier doped regime @xcite . these ladder cuprate materials also are natural extensions of the cu - o chain compounds towards the cuo@xmath4 sheet structures . within the framework of the kinetic - energy driven sc mechanism , some typical properties of the two - leg ladder cuprate superconductors have been studied , including the pressure dependence of @xmath2 @xcite , the charge dynamics @xcite , and the spin dynamics @xcite , and the calculated results are also qualitatively consistent with the experimental data obtained from the experimental measurements on the two - leg ladder cuprate superconductors . finally , we should be noted that much remains to be done . in particular , for the normal - state pseudogap , which grows upon underdoping , it seems natural to seek a connection to the physics of the af insulating parent compounds @xcite . however , at half - filling , the @xmath19-@xmath9 model is reduced as the af heisenberg model with an aflro . although a small density of charge carriers is sufficient to destroy aflro , this aflro remains until the extremely low - doped regime ( @xmath571 ) @xcite . as we have mentioned in eq . ( [ ecpair ] ) , the conduct is disrupted by aflro at the extremely low - doped regime , and then an important issue is how to extend the theory of the normal - state pseudogap state for the doped regime without aflro to the case at the extremely low - doped regime with aflro for a proper description of the connection between the finite doping normal - state pseudogap and the zero - doping quasiparticle dispersion . one of authors ( sf ) thanks professor z. b. su and professor l. yu for the early collaborations , and he also thanks li cheng , huaiming guo , zheyu huang , zhongbing huang , zhihao geng , mateusz krzyzosiak , ying liang , bin liu , tianxing ma , jihong qin , yun song , weifang wang , zhi wang , feng yuan , and jingge zhang for the collaborations . this work was supported by the funds from the ministry of science and technology of china under grant nos . 2011cb921700 and 2012cb821403 , the national natural science foundation of china under grant nos . 11274044 and 11447144 , and the science foundation of hengyang normal university under grant no . the charge - carrier operators @xmath295 and @xmath104 in the basis @xcite , @xmath572 of the charge - carrier states are given by , @xmath573 while the spin raising and lowering operators @xmath79 and @xmath78 in the spin 1/2 space , @xmath574 are given by , @xmath575 in the product space @xmath122 , the basis vectors are @xcite , however , as we have mentioned in eq . ( [ rhspace ] ) , the restricted hilbert space without double electron occupancy in the @xmath19-@xmath9 model ( [ tjmodel ] ) consists of three states , @xmath119 , @xmath120 , @xmath121 , namely , @xmath578 to remove the extra degrees of freedom in the @xmath122 space , we @xcite introduce a projection operator @xmath127 . by requiring @xmath579 , @xmath580 , and @xmath581 , we can easily obtain its matrix representation , @xmath582 and its hermitian conjugation , @xmath583 using this projection operator , the electron operators in the restricted hilbert space without double electron occupancy are given by @xcite , as quoted in eq . ( [ cep2 ] ) . it is then straightforward to verify the operator relations quoted in eq . ( [ sr3 ] ) . in particular , the charge - carrier number operator , @xmath585 the physical meaning of eq . ( [ hole - number ] ) is transparent : the empty state should be counted only once , not twice . since the mf treatment of the constraint on average doping concentration @xmath40 is imposed directly on @xmath586 , the sum rule for the constrained electron is satisfied . p. w. anderson , science * 235 * , 1196 ( 1987 ) ; p. w. anderson , in _ frontiers and borderlines in many particle physics _ , edited by r. a. broglia and j. r. schrieffer ( north - holland , amsterdam , 1987 ) , p. 1 . see , e.g. , the review , masaki fujita , haruhiro hiraka , masaaki matsuda , masato matsuura , john m. tranquada , shuichi wakimoto , guangyong xu , and kazuyoshi yamada , j. phys . soc . jpan . * 81 * , 011007 ( 2012 ) . a. g. loeser , z .- x . shen , d. s. dessau , d. s. marshall , c. h. park , p. fournier , and a. kapitulnik , science * 273 * , 325 ( 1996 ) ; h. ding , t. yokoya , j. c. campuzano , t. takahashi , m. randeria , m. r. norman , t. mochiku , k. kadowaki and j. giapintzakis , nature * 382 * , 51 ( 1996 ) . see , e.g. , the review , j. c. campuzano , m. r. norman , m. randeira , in _ physics of superconductors _ , vol . ii , edited by k. h. bennemann and j. b. ketterson ( springer , berlin heidelberg new york , 2004 ) , p. 167 . s. sorella , g. b. martins , f. becca , c. gazza , l. capriotti , a. parola , and e. dagotto , phys . lett . * 88 * , 117002 ( 2002 ) ; l. spanu , m. lugas , f. becca , and s. sorella , phys . rev . b * 77 * , 024510 ( 2008 ) . f. carbone , a. b. kuzmenko , h. j. a. molegraaf , e. van heumen , v. lukovac , f. marsiglio , d. van der marel , k. haule , g. kotliar , h. berger , s. courjault , p. h. kes , and m. li , phys . b * 74 * , 064510 ( 2006 ) . y. j. uemura , g. m. luke , b. j. sternlieb , j. h. brewer , j. f. carolan , w. n. hardy , r. kadono , j. r. kempton , r. f. kiefl , s. r. kreitzman , p. mulhern , t. m. riseman , d. ll . williams , b. x. yang , s. uchida , h. takagi , j. gopalakrishnan , a. w. sleight , m. a. subramanian , c. l. chien , m. z. cieplak , gang xiao , v. y. lee , b. w. statt , c. e. stronach , w. j. kossler , and x. h. yu , phys . rev . * 62 * , 2317 ( 1989 ) ; y. j. uemura , l. p. le , g. m. luke , b. j. sternlieb , w. d. wu , j. h. brewer , t. m. riseman , c. l. seaman , m. b. maple , m. ishikawa , d. g. hinks , j. d. jorgensen , g. saito , and h. yamochi , phys . * 66 * , 2665 ( 1991 ) . d. l. feng , d. h. lu , k. m. shen , c. kim , h. eisaki , a. damascelli , r. yoshizaki , j .- shimoyama , k. kishio , g. d. gu , s. oh , a. andrus , j. odonnell , j. n. eckstein , and z .- x . shen , science * 289 * , 277 ( 2000 ) . j. e. hirsch and f. marsiglio , phys . b * 45 * , 4807 ( 1992 ) ; d. j. scalapino , s. r. white , and s. c. zhang , phys . 68 * , 2830 ( 1992 ) ; d. j. scalapino , s. r. white , and s. c. zhang , phys . rev . b * 47 * , 7995 ( 1993 ) ; t. kostyrko , r. micnas , and k. a. chao , phys . rev . 6158 ( 1994 ) . ch . niedermayer , c. bernhard , u. binninger , h. glckler , j. l. tallon , e. j. ansaldo , and j. i. budnick , phys . rev . lett . * 71 * , 1764 ( 1993 ) ; l. fbrega , a. calleja , a. sin , s. piol , x. obradors , j. fontcuberta , and p. j. c. king , phys . rev . b * 60 * , 7579 ( 1999 ) ; j. l. tallon , j. w. loram , j. r. cooper , c. panagopoulos , and c. bernhard , phys . b * 68 * , 180501(r ) ( 2003 ) . y. j. uemura , a. keren , l. p. le , g. m. luke , w. d. wu , y. kubo , t. manako , y. shimakawa , m. subramanian , j. l. cobb , and j. t. markert , nature * 364 * , 605 ( 1993 ) ; m. niderst , r. frassanito , m. saalfrank , a. c. mota , g. blatter , v. n. zavaritsky , t. w. li , and p. h. kes , phys . . lett . * 81 * , 3231 ( 1998 ) ; s. l. lee , p. zimmermann , h. keller , m. warden , i. m. savi , r. schauwecker , d. zech , r. cubitt , e. m. forgan , p. h. kes , t. w. li , a. a. menovsky , and z. tarnawski , phys . lett . * 71 * , 3862 ( 1993 ) . m. le tacon , g. ghiringhelli , j. chaloupka , m. moretti sala , v. hinkov , m. w. haverkort , m. minola , m. bakr , k. j. zhou , s. blanco - canosa , c. monney , y. t. song , g. l. sun , c. t. lin , g. m. de luca , m. salluzzo , g. khaliullin , t. schmitt , l. braicovich , and b. keimer , nature phys . * 7 * , 725 ( 2011 ) . m. p. m. dean , g. dellea , r. s. springell , f. yakhou - harris , k. kummer , n. b. brookes , x. liu , y .- sun , j. strle , t. schmitt , l. braicovich , g. ghiringhelli , i. boovi , and j. p. hill , nature materials * 12 * , 1019 ( 2013 ) . m. p. m. dean , a. j. a. james , r. s. springell , x. liu , c. monney , k. j. zhou , r. m. konik , j. s. wen , z. j. xu , g. d. gu , v. n. strocov , t. schmitt , and j. p. hill , phys . rev . lett . * 110 * , 147001 ( 2013 ) . m. le tacon , m. minola , d. c. peets , m. moretti sala , s. blanco - canosa , v. hinkov , r. liang , d. a. bonn , w. n. hardy , c. t. lin , t. schmitt , l. braicovich , g. ghiringhelli , and b. keimer , phys . b * 88 * , 020501(r ) ( 2013 ) . l. braicovich , j. van den brink , v. bisogni , m. moretti sala , l. j. p. ament , n. b. brookes , g. m. de luca , m. salluzzo , t. schmitt , v. n. strocov , and g. ghiringhelli , phys . 104 * , 077002 ( 2010 ) . m. guarise , b. dalla piazza , m. moretti sala , g. ghiringhelli , l. braicovich , h. berger , j. n. hancock , d. van der marel , t. schmitt , v. n. strocov , l. j. p. ament , j. van den brink , p .- h . lin , p. xu , h. m. rnnow , and m. grioni , phys . 105 * , 157006 ( 2010 ) . b. j. sternlieb , g. shirane , j. m. tranquada , m. sato , and s. shamoto , phys . b * 47 * , 5320 ( 1993 ) ; r. j. birgeneau , r. w. erwin , p. g. gehring , m. a. kastner , b. keimer , m. sato , s. shamoto , g. shirane , and j. tranquada , z. phys . b * 87 * , 15 ( 1992 ) . a. v. puchkov , p. fournier , t. timusk , and n. n. kolesnikov , phys . lett . * 77 * , 1853 ( 1996 ) ; a. v. puchkov , p. fournier , d. n. basov , t. timusk , a. kapitulnik , and n. n. kolesnikov , phys . lett . * 77 * , 3212 ( 1996 ) . seyed iman mirzaei , damien stricker , jason n. hancock , christophe berthod , antoine georges , erik van heumen , mun k. chan , xudong zhao , yuan li , martin greven , neven barii , and dirk van der marel , pnas * 110 * , 5774 ( 2013 ) . j. w. loram , k. a. mirza , j. r. cooper , w. y. liang , and j. m. wade , j. supercond . * 7 * , 243 ( 1994 ) ; j. w. loram , k. a. mirza , j. m. wade , j. r. cooper , and w. y. liang , physica c * 235 - 240 * , 134 ( 1994 ) ; j. w. loram , k. a. mirza , j. r. cooper , and j. l. tallon , physica c * 282 - 287 * , 1405 ( 1997 ) . w. y. liang , j. w. loram , k. a. mirza , n. athanassopoulou , and j. r. cooper , physica c * 263 * , 277 ( 1996 ) ; j. w. loram , j. l. luo , j. r. cooper , w. y. liang , and j. l. tallon , physica c * 341 - 348 * , 831 ( 2000 ) .	superconductivity in cuprate superconductors occurs upon charge - carrier doping mott insulators , where a central question is what mechanism causes the loss of electrical resistance below the superconducting transition temperature ? in this review , we attempt to summarize the basic idea of the kinetic - energy driven superconducting mechanism in the description of superconductivity in cuprate superconductors . the mechanism of the kinetic - energy driven superconductivity is purely electronic without phonons , where the charge - carrier pairing interaction in the particle - particle channel arises directly from the kinetic energy by the exchange of spin excitations in the higher powers of the doping concentration . this kinetic - energy driven d - wave superconducting - state is controlled by both the superconducting gap and quasiparticle coherence , which leads to that the maximal superconducting transition temperature occurs around the optimal doping , and then decreases in both the underdoped and overdoped regimes . in particular , the same charge - carrier interaction mediated by spin excitations that induces the superconducting - state in the particle - particle channel also generates the normal - state pseudogap state in the particle - hole channel . the normal - state pseudogap crossover temperature is much larger than the superconducting transition temperature in the underdoped and optimally doped regimes , and then monotonically decreases upon the increase of doping , eventually disappearing together with superconductivity at the end of the superconducting dome . this kinetic - energy driven superconducting mechanism also indicates that the strong electron correlation favors superconductivity , since the main ingredient is identified into a charge - carrier pairing mechanism not from the external degree of freedom such as the phonon but rather solely from the internal spin degree of freedom of the electron . the typical properties of cuprate superconductors discussed within the framework of the kinetic - energy driven superconducting mechanism are also reviewed .
the x - ray light curves of numerous accreting neutron star and stellar - mass black holes ( stmbhs ) show evidence for the presence of quasi - periodic oscillations ( qpos ) , which appear as finite - width peaks in their power density spectra ( pds ) ( see van der klis 2006 and mcclintock & remillard 2006 for reviews of neutron star and stmbh qpos ) . while it is known that qpos occur with a wide range of centroid frequencies a few mhz to above a khz in neutron stars and a few mhz to a few hundred hz in the case of stmbhs the exact nature of the physical processes producing such oscillations is still a mystery . based on the observed properties , that is , their centroid frequencies , widths , amplitudes , and overall nature of their power spectra , etc . , qpos have been categorized into different groups . in neutron star binaries the qpo phenomenon constitutes the _ kilohertz _ qpos ( centroid frequencies in the range of 300 - 1200 hz : see the review by van der klis 2000 ) seen from over two dozen sources ( e.g. , mndez et al . 2001 ; barret et al . 2008 and references therein ) , the _ hectohertz _ qpos ( @xmath9100 - 300 hz : e.g. , van straaten et al . 2003 ; altamirano et al . 2008a ) seen predominantly in a special class ( atoll ) of neutron star binaries , the low - frequency qpos ( 0.01 - 50 hz : e.g. , van straaten et al . 2003 ) , the 1 hz qpos observed in two accreting millisecond x - ray pulsars ( e.g. , wijnands 2004 ) , the @xmath10 hz qpos observed only from dipping ( high - inclination ) neutron star binaries ( e.g. , homan et al . 1999 ; jonker et al . 1999 , 2000 ) and the very low - frequency 7 - 15 mhz qpos observed from at least three systems ( e.g. , revnivtsev et al . 2001 , altamirano et al . 2008b ) . similarly , black holes also show a variety of qpos ( mcclintock & remillard 2006 ) . they can be broadly classified into two categories : ( 1 ) high - frequency qpos ( hfqpos ) with centroid frequencies in the range of a few@xmath11(10 - 100 ) hz ( e.g. , miller et al . 2001 ; strohmayer 2001 ; remillard et al . 2006 ; belloni & altamirano 2013 ) and ( 2 ) low - frequency qpos ( lfqpos ) that occur in the range of 0.1 - 15 hz ( e.g. , casella et al . 2005 ) . based on their broadband properties , viz . , shape , fractional amplitude of the pds and the qpos , etc . , the lfqpos have been further sub - divided into type - a , b and c ( e.g. , homan et al . 2001 ; remillard et al . in addition to the hfqpos and the lfqpos of stmbhs , two black hole sources , grs 1915 + 105 and igr j17091@xmath83624 , show so - called `` heartbeat '' qpos which occur in the mhz frequency regime ( e.g. , belloni et al . 2000 ; altamirano et al . 2011 ) . furthermore , some ultraluminous x - ray sources ( ulxs ) show a few@xmath1110 mhz qpos ( e.g. , dewangan et al . 2006 ; pasham & strohmayer 2013 ) . more recently , an 11 mhz x - ray qpo and the recurrence of a few@xmath11mhz qpos were detected from the black hole candidates h1743@xmath8322 and cygnus x-3 , respectively ( koljonen et al . 2011 ; altamirano & strohmayer 2012 ) . ic 10 x-1 is an eclipsing , wolf - rayet binary containing the most massive stmbh known with an estimated black hole mass of 23 - 34 m@xmath12 ( prestwich et al . 2007 ; silverman & filippenko 2008 ) . the presence of an eclipse suggests that the system is highly inclined , i.e. , close to edge - on . this source was observed previously with _ xmm - newton _ ( i d : 0152260101 ) for a duration of roughly 45 ks . after accounting for background flaring only a mere 15 ks of useful data was available , analysis of which showed some evidence although at modest statistical significance for the presence of a qpo at @xmath0 7 mhz . motivated by this , and to carry out eclipse mapping , a long _ xmm - newton _ observation was proposed to confirm the presence of this mhz qpo ( i d : 0693390101 ; pi : strohmayer ) . here we present results from our timing analysis of this new data set and confirm the presence of the qpo at 7 mhz . beginning 2012 august 18 at 22:05:46 ( utc ) _ xmm - newton _ observed ic 10 x-1 for roughly 135 ks , a duration approximately equal to its orbital period ( 34.93 hrs : prestwich et al . 2007 ; silverman & filippenko 2008 ) . for our study we only used the epic data ( both pn and mos ) . we used the latest standard analysis system version 13.0.0 for extracting the images and the event lists . the source was easily identifiable and there were no source confusion problems ( see figure 1 ) . the source events were extracted from a circular region of radius 33 centered around the source and the background events were extracted from a nearby circular region of the same size . the observation was affected by flaring only briefly at the very beginning and the end of the pointing . these epochs were removed from our analysis . the combined pn and mos 0.3 - 10.0 kev light curve of ic 10 x-1 is shown in figure 2 ( _ black _ ) along with the background ( _ red _ ) . also overlaid are the good time intervals ( gtis ) during which a given epic instrument ( pn / mos1/mos2 ) was continuously active for more than 5 ks . for a given instrument the horizontal line , which is offset to an arbitrary value , indicates the active time , while a vertical line marks the beginning or the end of a continuous gti . it is clear that epic - pn has three gtis of duration roughly 23 ks , 75 ks and 27 ks , while mos1 has two gtis of length 30 ks and 99 ks and mos2 has one long gti of 130 ks . it is clear even by eye that the source varies significantly . since the pn detector offers the highest effective area among the three epic instruments , we started our analysis with its longest available gti of @xmath0 75 ks ( between hour 8.5 to 29.5 in figure 2 ) . using all the 0.3 - 10.0 kev photons we constructed a leahy - normalized pds where the poisson noise equals 2 ( leahy et al . this is shown in the top left panel of figure 3 ( _ histogram _ ) . it is evident that the overall power spectrum can be described by a simple power - law noise at the lowest frequencies with a qpo - like feature around 7 mhz and essentially poisson noise at frequencies above @xmath7 0.02 hz . in order to test the significance of the qpo we followed a rigorous monte carlo approach described below . first , we fit the continuum of the pds using a model consisting of a power - law plus a constant . while modeling the continuum we used the frequency range 0.0001 ( the lowest that can be probed ) - 0.1 hz and excluded the region containing the apparent qpo feature , i.e. , 5 - 9 mhz . the best - fit continuum model parameters are shown in the first column of table 1 . thereafter , following the prescription described by timmer & koenig ( 1995 ) , we simulated a large number of light curves ( and their corresponding pds ) that have the same shape , i.e. , same parameters , as in the 1@xmath13 column of table 1 , and the same frequency resolution as the spectrum used for obtaining the best - fit continuum parameters . a sample pds simulated with the above technique ( _ red _ ) along with the real pds ( _ black _ ) is shown in the top right panel of figure 3 . we simulated 370 such power spectra and found the maximum value in each frequency bin . this gave us the 99.73% ( 3@xmath2 ) significance within that particular bin . a similar estimate for each frequency bin gave us the complete confidence curve . similarly we simulated 10000 pds and estimated the 99.99% ( 3.9@xmath2 ) confidence curve . it should be noted that the confidence curves are sensitive to the chosen values of the continuum model parameters . given the error on each of the individual model parameters , i.e. , best - fit power - law normalization and the power - law index , we estimated the 99.73% and the 99.99% curves for various combinations of the power - law normalization and the index within the error bars quoted in column 1 of table 1 . to be conservative we picked the maximum of these curves . these confidence levels are overlaid in the figure . it is clear that the qpo feature is significant at the 99.99% level . to further confirm the presence of this qpo feature we extracted another pds using the combined mos data . for this purpose we used the longer gti of roughly 95 ks ( from hour 8.5 to hour 34.5 in figure 2 ) . the 0.3 - 10.0 kev combined mos pds is shown in the bottom left panel of figure 3 . the qpo feature is again evident at 7 mhz . to quantify the variability we first modeled the pds with a power - law plus a constant this gave a @xmath14 of 245 with 192 degrees of freedom ( dof ) . we then added a lorentzian component to model the qpo feature at 7 mhz . this improved the @xmath14 by 28 with an addition of three parameters , i.e. , a @xmath14 of 217 with 189 dof ( see the second column of table 1 ) . this decrease in @xmath14 serves as a further indicator of the significance of the qpo component . using the f - test this corresponds to a single - trial significance of 4@xmath1110@xmath15 . note that we essentially searched in a known narrow frequency range of @xmath0 5 - 10 mhz ( from the prior _ xmm - newton _ observation ) , and thus the effective number of trials is @xmath9 1 . it is clear that the qpo is present in two independent detectors at @xmath1 3.5@xmath2 confidence level in each case . the chance probability of two independent 3@xmath2 detections alone is 4.33@xmath2 . given the 3.5@xmath2 detections in two separate measurements , we conclude that the observed 7 mhz qpo of ic 10 x-1 is statistically highly significant . in addition , we studied the energy dependence of the fractional rms amplitude of the qpo . for this purpose , we extracted the pds of the source in seven energy bands . owing to the low count rate we fixed the upper bound of the bandpass at 10 kev and varied the lower limit from 0.3 to 1.5 kev and constructed a pds in each case using the combined mos data . each of these pds were then modeled with a power - law plus constant for the continuum and a lorentzian for the qpo ( best - fit model parameters shown in table 1 ) . the fractional rms amplitude of the qpo is : @xmath16 where @xmath17 and @xmath18 are the normalization and the width of the qpo ( lorentzian ) , respectively , while @xmath19 is the mean count rate of the source . the dependence of the rms amplitude of the qpo as a function of the lower limit of the band pass is shown in the bottom right panel of figure 3 . there is a very weak dependence of the qpo s amplitude on the energy from 0.3 - 1.5 kev . four of the five pds used for this analysis are shown in figure 4 . at energies greater than 1.5 kev the low signal to noise ratio of the data does not allow us to detect the qpo . numerous x - ray binaries and also active galactic nuclei show evidence for the presence of a break in their pds ( e.g. , mchardy et al . 2006 ; markowitz & edelson 2004 ) . we searched for a spectral break in ic 10 x-1 using the data from the longest gti outside the eclipse , i.e. , hour 15.5 to 33.5 in figure 2 . note that the presence of an eclipse in the data adds red noise to the power spectrum that is not intrinsic to the source variability . we constructed the combined mos 0.3 - 10.0 kev pds and did not find any obvious evidence for a pds break down to frequencies as low as 0.0001 hz . note that a single pds is noisy , with error in a particular bin equal to the value of that bin ( van der klis 1989 ) . therefore , averaging ( say , by combining neighboring bins ) is necessary to reduce the noise in the pds . hence , even though the lowest sampled frequency is @xmath0 10@xmath15 hz ( 1/total length ) averaging reduces the lowest effective frequency to roughly 0.0001 hz in this case . it remains possible that a break may exist at @xmath20 0.0001 hz . moreover , we modeled this pds with a power - law and a lorentzian . we find that the best - fit continuum can be described by a power - law of index @xmath6 . the frequency of the hfqpos of stmbhs ( @xmath9 a few 100 hz ) and the _ hectohertz _ qpos of neutron stars ( @xmath9 100 - 300 hz ) are roughly constant in frequency for a given source ( van der klis 2006 ) . they are thought to have a common origin ( e.g. , abramowicz et al . 2003 ) and it has been proposed that the qpo frequency may scale inversely with the mass of the compact object ( e.g. , see figure 4.17 of mcclintock & remillard 2006 ) . with a mass of 23 - 34 m@xmath12 ic 10 x-1 s hfqpos , if any , are expected to occur in the range of a few 10s of hz . clearly the 7 mhz qpo of ic 10 x-1 is orders of magnitude slower than this and is very likely not a hfqpo phenomenon . the typical values of the centroid frequency , rms amplitude and the quality factor ( @xmath21 centroid - frequency / qpo - width ) of type - a lfqpos are @xmath22 hz , @xmath23 and @xmath23 , respectively ( e.g. , casella et al . 2005 ) . the respective values for type - b lfqpos are @xmath24 hz , @xmath25 and @xmath26 . lastly , type - c qpos occur in a wider range of frequencies 0.1 - 15 hz and have rms amplitudes of 3 - 20 with @xmath27 factors of @xmath28 . the overall nature of the pds accompanying type - a , b , and c qpos can be described as weak red noise , weak red noise and strong flat - topped noise , respectively . although the continuum of the pds of ic 10 x-1 is similar to that accompanying type - a or b qpos , its qpo frequency , rms amplitude and @xmath27 value are quite different ( compare values in table 1 with table 1 of casella et al . on the other hand , the centroid frequency of ic 10 x-1 s qpo is much lower compared to a typical type - c qpo . the mhz qpos ( frequency range of @xmath9 10 - 200 mhz ) of ulxs have been argued to be the analogs of the type - c lfqpos of stmbhs but occurring at a lower frequency ( a few 10s of mhz compared to the few hz of stmbhs ) due to the presence of intermediate - mass black holes ( mass of a few@xmath11(100 - 1000 ) m@xmath12 ) within these systems . while the centroid frequency , the rms amplitude and the @xmath27 value of the 7 mhz qpo of ic 10 x-1 are comparable to the mhz qpos of ulxs ( e.g. , dheeraj & strohmayer 2012 ) , there are two aspects that are dissimilar . ( 1 ) we do not detect a break in the pds of ic 10 x-1 whereas breaks have been seen in all the ulxs ( e.g. , dewangan et al . it is known that the break frequency scales with the qpo frequency as @xmath29 @xmath9 @xmath30/9 ( wijnands & van der klis 1999 ) . if that were the case for ic 10 x-1 , the expected break is at @xmath9 0.7 mhz . it is thus possible that we are unable to detect the break due to our inability to sample variability at very low ( @xmath20 0.7 mhz ) frequencies , or the effects of the eclipse . ( 2 ) the rms amplitude at least in the case of the ulx ngc 5408 x-1is known to increase with energy from 0.3 - 2.0 kev ( strohmayer et al . 2007 ; middleton et al . 2011 ) . however , we do not find evidence for such behavior in ic 10 x-1 ( see the right panel of figure 3 ) . two black hole sources , grs 1915 + 105 and igr j17091@xmath83624 , show mhz qpos in the so - called `` heartbeat '' state or the @xmath31 state ( e.g. , greiner et al . 1996 ; morgan et al . 1997 ; belloni et al . 2000 ; altamirano et al . 2011 ) . these qpos are thought to be the result of a radiation pressure instability causing quasi - periodic evaporation followed by refilling of the inner regions of the accretion disk ( lightman & eardley 1974 , belloni et al . 1997 ; neilsen 2011 ) . these mhz qpos occur at relatively high luminosities ( @xmath9 10@xmath32 erg s@xmath33 ) and at least in grs 1915 + 105 appear to be energy - independent in the band pass from 2 - 30 kev ( see figure 8 of morgan et al . given the similar frequency , comparable rms amplitude ( e.g. , altamirano et al . 2011 ) , energy independence ( although here energy independence is seen over a different x - ray band pass ) and similar x - ray luminosity 0.9 kev . this model gives an acceptable fit with a @xmath14 of 165 for 125 degrees of freedom . a detailed spectral analysis is not the subject of this work . nevertheless , assuming this simple model , the inferred 2 - 10 kev luminosity at a distance of 0.66 mpc is @xmath9 10@xmath32 ergs s@xmath33 . ] ( see also wang et al . 2005 ) it is possible that the 7 mhz qpo of ic 10 x-1 is related to the `` heartbeat '' qpos . given its low count rate it is , however , not possible to resolve ic 10 x-1 s light curve to the same level as grs 1915 + 105 or igr j17091@xmath83624 . it will require instruments with larger collecting area to test this hypothesis . the 7 mhz qpo of ic 10 x-1 is likely not related to the 1 hz qpos of accreting millisecond x - ray pulsars ( amxps ) or the 7 - 15 mhz qpos seen in some neutron star systems ( e.g. , revnivtsev et al . 2001 , altamirano et al . the 1 hz qpo phenomenon in amxps is thought to be due to disk instabilities within the boundary layer of the accretion disk and neutron star magnetosphere ( patruno et al . 2009 ) . moreover , the 1 hz qpos are seen at low luminosities ( @xmath34 10@xmath35 erg s@xmath33 ) and can have rms amplitudes as large as 100% ( patruno et al . 2009 ) . given the high rms amplitude , and that they are likely related to the beginning of the propeller regime , they are probably different from the qpo seen in ic 10 x-1 . the 7 - 15 mhz oscillations in some neutron stars are linked to marginally stable burning of light elements on the surface of the neutron star ( e.g. , heger et al . 2007 ) , a process unique to neutron stars . on the other hand , the 7 mhz qpo may be connected to the 0.6 - 2.4 hz `` dipper qpos '' of high - inclination neutron stars in the sense that ic 10 x-1 is also highly inclined ( see the eclipse in figure 1 and silverman & filippenko 2008 ) . the so - called `` dipper qpos '' are only seen from x - ray dipping sources . the dipping is presumably due to obscuration associated with the high inclination ( parmar & white 1988 ) . their rms amplitudes are @xmath9 10% and are energy - independent in the range of 2 - 30 kev ( see , for example , figure 4 of homan et al . 1999 ) . with the present _ xmm - newton _ data a similar energy range can not be probed . however , we note that the qpo s rms amplitude is comparable to those of `` dipper qpos '' and appears to be independent of energy in the range from 0.3 - 1.5 kev . more recently altamirano & strohmayer ( 2012 ) reported the discovery of an 11 mhz qpo from a black hole candidate h1743 - 322 ( likely highly inclined : homan et al . 2005 ) and suggested this could be the first detection of a `` dipper qpo '' analog in a black hole ( candidate ) system . the centroid frequency of the `` dipper qpos '' is roughly constant over time . if the 7 mhz qpo is indeed a `` dipper qpo '' then it s centroid frequency should also remain more or less constant . this can be tested with multi - epoch observations of ic 10 x-1 to search for qpo variability . finally , we note that the overall pds and the qpo properties of ic 10 x-1 are also similar to that of cygnus x-3 ( van der klis & jansen 1985 ) . they both have the same power - law like noise at low frequencies each with roughly the same slope of @xmath36with a qpo in the mhz regime and barely any power above 0.1 hz ( see the bottom panels of figure 2 of koljonen et al . in addition the qpos in both cases have comparable frequencies , rms amplitudes and coherences ( van der klis & jansen 1985 ) . the mhz oscillations of cygnus x-3 are likely associated with major radio flaring events ( koljonen et al . 2011 ; see , for example , lozinskaya & moiseev ( 2007 ) and references therein for radio studies of ic 10 x-1 ) . simultaneous radio and x - ray observations in the future can test whether jet ejection is related to the qpo in ic 10 x-1 .	we report the discovery with _ xmm - newton _ of an @xmath0 7 mhz x - ray ( 0.3 - 10.0 kev ) quasi - periodic oscillation ( qpo ) from the eclipsing , high - inclination black hole binary ic 10 x-1 . the qpo is significant at @xmath1 4.33@xmath2 confidence level and has a fractional amplitude ( % rms ) and a quality factor , @xmath3 , of @xmath4 and @xmath5 , respectively . the overall x - ray ( 0.3 - 10.0 kev ) power spectrum in the frequency range 0.0001 - 0.1 hz can be described by a power - law with an index of @xmath6 , and a qpo at 7 mhz . at frequencies @xmath7 0.02 hz there is no evidence for significant variability . the fractional amplitude ( rms ) of the qpo is roughly energy - independent in the energy range of 0.3 - 1.5 kev . above 1.5 kev the low signal to noise ratio of the data does not allow us to detect the qpo . by directly comparing these properties with the wide range of qpos currently known from accreting black hole and neutron stars , we suggest that the 7 mhz qpo of ic 10 x-1 may be linked to one of the following three categories of qpos : ( 1 ) the `` heartbeat '' mhz qpos of the black hole sources grs 1915 + 105 and igr j17091@xmath83624 , or ( 2 ) the 0.6 - 2.4 hz `` dipper qpos '' of high - inclination neutron star systems , or ( 3 ) the mhz qpos of cygnus x-3 .
the quantum mechanical effects resulting from time - dependent boundary conditions have been studied by some authors @xcite . diffraction in time was initially introduced by moshinsky @xcite by considering a situation involving a beam of particles impinging from the left on a totally absorbing shutter located at the origin which is suddenly turned off at an instant . in such an example , the transient probability current has a close mathematical resemblance with the intensity of light resulting from the fresnel diffraction by a straight edge . see @xcite for a recent review . an interesting feature of the solutions for cutoff initial waves , occurring both in the free case @xcite and in the presence of a potential interaction @xcite , is that , if initially there is a zero probability for the particle to be at @xmath3 , as soon as @xmath4 , there is instantaneously , a finite , though very small , probability to find the particle at any point @xmath3 . this non - local behavior of the schrdinger solution is due to its nonrelativistic nature and not a result of the quantum shutter setup @xcite . application of the klein@xmath5gordon equation to the shutter problem @xcite shows that the probability density is restricted to the accessible region @xmath6 ( @xmath7 is the speed of light ) . gerasimov and kazarnovskii @xcite confined the initial wave in a finite region by introducing a second shutter at the point @xmath8 . godoy @xcite pointed out the analogy with fraunhoffer diffraction in the case of small box ( compared to the de broglie length ) , and fresnel diffraction , for larger confinements . on the other hand , the phenomenon of quantum superarrivals for the reflection and transmission of a gaussian wave packet in the case of a rectangular potential barrier while it is perturbed by either reducing or increasing its height has also been considered @xcite . there exists a finite time interval during which the probability of reflection is larger ( superarrivals ) when the barrier is lowered compared to the unperturbed case @xcite . similarly , during a certain interval of time , the probability of transmission when the barrier is raised exceeds that for free propagation @xcite . such studies suggest that the time - varying boundary conditions can give rise to interesting action - at - a - distance effects in quantum mechanics . exactly solvable models give important points about the quantum many - body physics . the thonks - girardeau ( tg ) gas @xcite is a such model , which corresponds to infinitely repulsive bosons in one dimension ( 1d ) . this model is exactly solvable via a fermi - bose mapping , which relates the tg gas to a system of noninteracting spinless fermions in 1d @xcite . in this context , by considering the problem of a confined particle in a linear potential , the aim of the present paper is to probe some aspects of the time - dependent boundary conditions that have remained hitherto unnoticed . the organization of the paper is as follows : in section [ sec : review ] we review the solution of the schrdinger equation for a confined particle in half space @xmath9 with the potential @xmath1 and consider the various time - dependent boundary condition of the problem . section [ sec : qu_st ] briefly deals with quantum statistics . numerical calculations are presented in section [ sec : nuca ] . section [ sec : sudi ] contains concluding remarks . the schrdinger equation of the @xmath10 state of the particle wave function is @xmath11 performing the change of variable @xmath12 we get the airy equation , @xmath13 whit solutions , @xmath14 in which @xmath15 is the normalization constant . the energy levels are determined by the condition @xmath16 @xmath17 & = & 0 ~.\end{aligned}\ ] ] then , the @xmath18 are determined by the zeros @xmath19 of the airy function @xmath20 thus , the solution of eq . ( [ eq : sch ] ) is given by @xcite , @xmath21 in which @xmath22 . now , we study the evolution of eq . ( [ eq : initial_wave ] ) under time - dependent schrdinger equation in the following cases : suddenly the wall at @xmath23 is removed at time equal to zero , and a linear potential of the form @xmath1 is switched on in the region @xmath2 . now , the particle is in a linear potential in whole space . at any instant @xmath24 wavefunction is obtained by , @xmath25 in which @xmath26 is the initial wavefunction and is chosen as eq . ( [ eq : initial_wave ] ) and @xmath27 is the propagator for the hamiltonian @xmath28 and is given by @xcite , @xmath29 putting eq . ( [ eq : prop - li ] ) into eq . ( [ eq : wf(t ) ] ) , one finds @xmath30 inverse fourier transform of the function @xmath31 is the airy function , @xmath32 so , @xmath33~,\end{aligned}\ ] ] in which @xmath34 and erfc@xmath35 is the complementary error function . in this case the propagator is given by @xcite , @xmath41 where @xmath42 denotes the propagator of free particle and is given by eq . ( [ eq : prop - li ] ) with @xmath36 . from eq . ( [ eq : wf_fr ] ) , we have @xmath43 - ~\text{e}^{-ikx } \text{erfc } \left[\frac{i+1}{2 } \sqrt{\frac{m}{\hbar t } } ( -x - vt)\right ] \right)~.\end{aligned}\ ] ] so far we have not consider quantum statistics . we shall next consider dynamics of the ground state of a system of @xmath44 noninteracting spinless fermions initially confined in half - space @xmath0 in the presence of a linear potential . since wavefunctions of noninteracting spinless fermions are antisymmetric under coordinate exchanges , their wavefunctions vanish automatically whenever any @xmath45 : @xmath46 the single particle density , normalized to @xmath44 , is @xmath47 it follows from the fermi - bose mapping theorem @xcite that the exact @xmath48bosons ground state wavefunction @xmath49 of a system of @xmath44 bosons with hard - wall interaction is @xmath50 in this section we work in a unit system where @xmath51 and @xmath52 . we choose @xmath53 anywhere the potential is nonzero . thus , from eq . ( [ eq : initial_wave ] ) the initial wavefunction is given by , @xmath54 we use krger s method @xcite to produce airy functions . [ fig : denx_b ] shows the probability density versus distance @xmath55 for state @xmath56 with energy eigenvalue @xmath57 at different times . from this figure one can see that at long times probability density locates symmetrically around the origin and it spreads without motion . from eqs . ( [ eq : wf(t ) ] ) and ( [ eq : prop - li ] ) for free propagation one has , @xmath58 and @xmath59~,\end{aligned}\ ] ] in which c.c . refers to the complex conjugate of the first term in the bracket . one can check numerically in the limit @xmath60 , last line of the above expression is very small compared to the two first terms . as a result of which density probability would be an even function of @xmath55 at long times . after a certain time @xmath61 , for even @xmath62 , @xmath63 has a local minimum at @xmath23 and its @xmath62 maxima are distributed symmetrically around @xmath23 , but for odd @xmath62 it has a local maximum at @xmath23 and other @xmath64 maxima locate symmetrically around the central maximum , i.e. , @xmath65 maxima at the right and @xmath65 maxima at the left of the origin . another feature of @xmath63 is that it has no node after a certain time of letting the particle to be free . from eq . ( [ eq : free - linear ] ) one can find this is also true for propagating of @xmath66 in a linear potential . another physically interesting quantity is the probability current density @xmath67 . it is important because arrival time distribution at a given location is given by its modulus sign in some approaches to this problem @xcite . [ fig : curt_b ] shows the probability current density versus time at different locations . from this figure one can see @xmath68 is negative for negative values of @xmath55 and is positive for large positive values of @xmath55 , but its sign changes with time for small values of positive @xmath55 . [ fig : denprofile_b ] represents the evolution of the density profile for a noninteracting spinless fermioninc gas composed of @xmath69 atoms released from the trap . when the particle is trapped in the @xmath70th mode , and after being released , the probability density presents @xmath62 maxima and by elapsing the time , they distribute symmetrically around the origin . as a result of which , for the case of @xmath48body system , single particle density presents @xmath44 maxima , when particles are trapped and after being released and by passing the time these maxima locates symmetrically around the origin . fig . [ fig : denx_c ] shows probability density versus distance @xmath55 for state @xmath56 at different times . it cen be seen that wavefunction spreads without motion . the height of the last maximum is larger than the @xmath64 first ones before turning the potential off , but by going time , the height of them goes to be the same except the last one , whose height is smaller than the others . @xmath63 in fig . [ fig : denx_c ] has always @xmath62 nodes in spite of fig . [ fig : denx_b ] . this comes from the fact that in this case the wall is in place . [ fig : curt_c ] shows the probability current density versus time at different locations . from this figure one can see that @xmath68 is positive for large values of @xmath55 , but its sign changes with time for small values of @xmath55 . finally , fig . [ fig : denprofile_c ] represents the evolution of the density profile for a fermioninc gas composed of @xmath69 noninteracting spinless atoms released from the trap . noting this figure , one sees the height of the maxima decreases by @xmath55 for a given time . in this paper we have considered numerical solution of the time - dependent schrdinger equation for some time - dependent boundary conditions . probability density , probability current density and dynamics of the ground state of a system of @xmath44 noninteracting spinless fermions were studied . when the particle is trapped in the @xmath70th mode , and after being released , the probability density presents @xmath62 maxima . as a result of which , for the case of @xmath48body system , single particle density presents @xmath44 maxima , when particles are trapped and after being released . since the stationary wavefunctions @xmath66 given in eq . ( [ eq : initial_wave ] ) form a complete set of orthonormal eigenfunctions in the range @xmath71 , we can expand an initial wave packet @xmath72 which has an arbitrary shape for @xmath71 in terms of @xmath66 @xmath73 where the expansion coefficients @xmath74 is given by : @xmath75 . we have find time evolution of eigenfunctions @xmath66 which were denoted by @xmath76 . thus , time evolution of eq . ( [ eq : expansion ] ) is given by @xmath77 time - dependent probability density @xmath78 is a complicated linear combination of @xmath63 and the corresponding interferences amplitudes @xmath79 . 99 m. moshinsky , _ phys . * 88 * ( 1952 ) 625 a. s. gerasimov and m. v. kazarnovskii , _ sov . * 44 * ( 1976 ) 892 r. ghler and r. golub , _ z. phys . b _ * 56 * ( 1984 ) 5 a. del campo , g. garca - caldern and j g muga , _ phys . rep . _ * 1 * ( 2009 ) 476 p. r. holland , _ the quantum theory of motion _ ( cambridge university press , cambridge , 1993 ) g. garca - caldern , a. rubio , _ phys . rev . a _ * 55 * ( 1997 ) 3361 s. godoy , _ phys . a _ * 65 * ( 2002 ) 042111 s. bandyopadhyay , a. s. majumdar and d. home , _ phys.rev . a _ * 65 * ( 2002 ) 052718 md . m. ali , a. s. majumdar and d. home , _ a _ * 304 * ( 2002 ) 61 a. s. majumdar and d. home , _ pranama _ * 59 * ( 2002 ) 321 d. m. greenberger , _ physica b _ * 151 * ( 1988 ) 374 g. garca - caldern , a. rubio , j. villavicencio , _ phys . a _ * 59 * ( 1999 ) 1758 a. del campo , and j g muga , _ europhys . * 74 * ( 2006 ) 965 m. girardeau , _ j. math . phys . _ * 1 * ( 1960 ) 516 o. valle and m. soares , _ airy functions and applications to physics _ ( imperial college press , london , 2004 ) c. grosche and f. steiner , _ handbook of feynman path integrals ( springer tracts in modern physics _ vol 145 ) ( berlin : springer ) m. kleber , _ phys . * 236 * ( 1994 ) 331 m. d. girardeau , _ j. math . phys . _ * 1 * ( 1960 ) 516 m. d. girardeau , _ phys . * 139 * ( 1965 ) b500 m. d. girardeau and e. m. wright , _ phys . * 84 * ( 2000 ) 5691 h. krger , _ theo . . acta _ * 16 * ( 1980 ) 97 j. g. muga and c. r. leavens , _ phys . * 338 * ( 2000 ) 353	the solution of the time - dependent schrdinger equation is discussed for a particle confined in half - space @xmath0 with a linear potential @xmath1 in the following situations : ( a ) sudden removal of the wall and switching on the linear potential @xmath1 at @xmath2 , ( b ) sudden removal of the wall and the potential and ( c ) sudden removal of the potential . a brief discussion of the quantum statistic is presented .
the interpretation and theoretical description of measurements on quantum systems have been under debate since the birth of quantum theory @xcite . more recently , interest for this question has been revived by new developments in quantum information . quantum detection can be used either to extract information on quantum states or to monitor quantum systems ( quantum trajectories @xcite , quantum zeno effect @xcite ) . quantitative treatments of measurement models serve both to elucidate the self - consistency of quantum theory and its interpretation , and to calculate the time scales relevant for experiments . data for the decoherence time are in fact accumulating , in microwave cavities @xcite , in solid state devices like superconducting tunnel junction nanocircuits @xcite , and in electron beams interacting with a semiconducting plate @xcite . on a more fundamental ground , ever larger classes of nonlocal hidden - variable theories are being ruled out experimentally as competitors of quantum mechanics @xcite . we propose in this letter to extend the range of validity of the approach by zeh , zurek , and others based on environment - induced decoherence @xcite . in this vein , i.e. , on the basis of the probabilistic interpretation of quantum mechanics ( born rules ) and of a lack of knowledge about the microscopic degrees of freedom of the apparatus , we demonstrate that the reading of a macroscopic pointer of the apparatus reveals an eigenvalue of an observable of the measured quantum object , in spite of the unitary evolution of the composite system ( object and apparatus ) . we obtain explicit results for the object - pointer dynamics for a class of models under realistic assumptions . we would like to note that macroscopic manifestations of microscopic quantum fluctuations are not the exclusive privilege of measurement : they also appear e.g. in superfluorescence , where intense light pulses display substantial shot - to - shot fluctuations @xcite and in cosmic rays , where single particles propagating in a fluctuating spacetime generate large showers of particles in the atmosphere . during an ideal measurement the quantum object ( @xmath0 ) interacts with the pointer ( @xmath1 ) in such a way that a one - to - one correspondence arises between the eigenvalues of the measured object observable and macroscopically distinct pointer states . the coupling of @xmath1 with its environment ( `` bath '' @xmath2 ) causes decoherence . most previous work deals with these two interactions separately ( see , however , @xcite ) : a first step ( `` premeasurement '' ) exclusively treats the unitary evolution entangling @xmath0 and @xmath1 and yields a superposition of the macroscopically distinct object - pointer states associated with each eigenvalue of the measured observable . that latter `` schrdinger cat '' state is taken as the initial state for a second process , decoherence ; there , the quantum correlations between object and apparatus are transformed into classical correlations , as the superposition of object - pointer states is degraded to a statistical mixture of the same states . for such a sequential treatment to make physical sense , the time duration of the entanglement process would have to be short compared with the decoherence time @xmath3 . but since the latter tends to be very small for macroscopic superpositions , that assumption is quite questionable . a second shortcoming of some previous work lies in the assumption of initial statistical independence of @xmath1 and @xmath2 ; since these two systems can not be isolated from each other , a more realistic assumption is a metastable local thermal state of @xmath4 . as a third restriction , memory effects are often neglected for the ( reduced ) object - pointer dynamics . that markov approximation assumes @xmath5 to be larger than the bath correlation time @xmath6 , a condition not satisfied in some experiments @xcite and of questionable validity if macroscopic or even mesoscopic superpositions or mixtures arise during the object - pointer evolution . we overcome the three aforementioned deficiencies . the key is an assumption for a certain ordering of time scales : the decoherence time @xmath7 and the object - pointer interaction time @xmath8 must be small compared to the characteristic time @xmath9 ( @xmath10 ) of the evolution of the measured observable @xmath11 ( the pointer position @xmath12 ) under the object hamiltonian @xmath13 ( the pointer hamiltonian @xmath14 ) , @xmath15 in the last limit , @xmath16 denotes the inverse temperature of the bath . as desirable for a measurement , the free dynamics of @xmath0 then remains ineffective on @xmath11 during the measurement , i.e. , @xmath17 for @xmath18 . here @xmath19 means that the expectation value of @xmath20 in the object initial state is approximately equal to the @xmath21th moment of @xmath11 , for any @xmath22 . similarly , the free dynamics of @xmath1 is ineffective to move the pointer between @xmath23 and @xmath24 . note that @xmath25 if @xmath11 commutes with @xmath13 . the conditions on @xmath9 in ( [ timescales ] ) are necessary in order that an eigenstate @xmath26 of @xmath11 be left almost unchanged by the measurement . the conditions on @xmath10 ( in particular , the high - temperature limit @xmath27 ) would be difficult to avoid for a macroscopic pointer . a further key input is the quantum central limit theorem ( qclt ) @xcite which implies gaussian statistics ( wick theorem ) for the bath coupling agent in the pointer - bath interaction , as discussed below . the limits ( [ timescales ] ) and the qclt imply validity of our results for a wide class of objects and apparatus . an extreme case of universality @xcite arises under the further stipulation that entanglement and decoherence be faster than the decay of bath correlations ( time constant @xmath6 ) . indeed , for @xmath28 even the free bath motion remains ineffective during the measurement . opposite to that slow - bath limit is the markov regime @xmath29 . our analysis below covers both these regimes as well as the intermediate case @xmath30 . we consider a three - partite system comprising the microscopic object @xmath0 , a single - degree - of - freedom macroscopic pointer @xmath1 and a bath @xmath2 with many ( @xmath31 ) degrees of freedom ( labelled by @xmath32 ) . the following dynamical variables will come into play : for @xmath0 , the observable @xmath33 to be measured , assumed to have a discrete spectrum ; for @xmath1 , the position @xmath34 and momentum @xmath35 ; and for @xmath2 , a certain coupling agent @xmath36 . the pointer is coupled to @xmath0 and @xmath2 by the hamiltonians @xmath37 the object - pointer coupling hamiltonian @xmath38 is chosen so as to ( i ) not change the measured observable @xmath11 ( i.e. , @xmath39=0 $ ] ) ; ( ii ) be capable of shifting the pointer position by an amount proportional to @xmath33 , such that each eigenvalue @xmath40 of @xmath11 becomes tied up with a specific pointer reading ; ( iii ) involve a large coupling constant @xmath41 , so that different eigenvalues @xmath42 end up associated with pointer readings separated by large distances . the pointer - bath interaction is chosen for most efficient decoherence of distinct pointer positions @xcite ; the additivity of @xmath36 in contributions @xmath43 acting on single degrees of freedom of the bath will allow us to invoke the qclt . let us point out an essential difference between our model and the interacting spin model of @xcite . unlike in this reference , @xmath0 is strongly coupled to a single degree of freedom ( the pointer @xmath1 ) of the apparatus , e.g. with its total momentum @xmath44 in a given direction . the coupling of @xmath0 with the other apparatus degrees of freedom ( the bath @xmath2 , for us ) is assumed to be much weaker and can therefore be neglected ( see @xcite ) . the full hamiltonian is @xmath45 . we need not specify @xmath13 . the pointer hamiltonian @xmath46 must allow for a well - defined rest state . we assume that @xmath47 has a local minimum at @xmath48 , i.e. , @xmath49 and @xmath50 . the time scale for free pointer motion then is the period @xmath51 of oscillations around this minimum . like the coupling agent @xmath36 , the bath hamiltonian @xmath52 is assumed additive as @xmath53 with @xmath54\propto \delta_{\mu \nu } \dot{b}_\mu$ ] and @xmath55=0 $ ] , again clearing ground for the qclt . it is appropriate to require initial statistical independence of the object and the apparatus , and thermal equilibrium for the apparatus . the initial density operator @xmath56 of the object may represent a pure or a mixed state . the full initial density operator reads @xmath57 by invoking the high - temperature limit @xmath27 and the gaussian statistics of @xmath36 ( as implied by the qclt ) and by tracing out the bath we find ( see below ) the reduced density matrix @xmath58 of the pointer in the position representation , @xmath59 where @xmath60 is the thermal de broglie wavelength . the pointer potential here appears renormalized by the pointer - bath interaction as @xmath61 where @xmath62 is the autocorrelator of the bath coupling agent with respect to the free bath thermal state @xmath63 , @xmath64 and we assume @xmath65 . for stability of the whole apparatus the pointer - bath coupling must be weak enough so that @xmath66 ; we even bound the latter curvature finitely away from zero by , say , @xmath67 , i.e. , @xmath68 this makes sure that the initial density @xmath69 of pointer positions has a single peak at @xmath48 with a renormalized width @xmath70^{-1/2}$ ] of the order of the bare thermal fluctuation @xmath71 . if @xmath72 at large distances @xmath73 , the effective potential @xmath74 is unstable . the matrix elements ( [ eq - rhop(0 ) ] ) then correspond to ( the reduced pointer state of ) a _ local thermal equilibrium_. that local equilibrium for the apparatus can be achieved by preparing @xmath1 in some state localized near @xmath48 at time @xmath75 and then letting it interact with @xmath2 between @xmath75 and @xmath23 . if the thermalization time is small compared with the tunnelling escape time , one may choose @xmath76 larger than the former but much smaller than the latter time , so that @xmath1 is still within the effective potential well when the measurement starts at @xmath23 . in order to be able to prepare the apparatus in such a local equilibrium , the height @xmath77 of the two potential barriers surrounding the local minimum of @xmath74 at @xmath48 must be large compared with the thermal energy @xmath78 . thanks to ( [ eq - stability ] ) , this is the case provided that the bare potential @xmath47 satisfies the same requirement , i.e. , @xmath79 . interestingly , @xmath47 can be chosen such that the two potential barriers of @xmath74 are separated by a mesoscopic distance @xmath80 ( so that @xmath81 ) which is small compared with the macroscopic read - out scale @xmath82 . the object - pointer interaction then just has to get the pointer out of the well , leaving the subsequent displacement growth to the action of the effective pointer potential . the instability resulting from the pointer - bath coupling ( [ eq - int_hamiltonian ] ) hence provides an amplification mechanism . for a macroscopic pointer at high temperature ( @xmath27 ) , the different length scales are ordered as @xmath83 . assuming that the bath state is not ascertainable , we define the object - pointer density matrix at time @xmath84 as @xmath85 . accepting a relative error @xmath86 , we simplify the time evolution operator at time @xmath87 as @xmath88 , with @xmath89 here , @xmath90 denotes time ordering ; the momentum @xmath44 was used as the generator of pointer displacements , @xmath91 . similarly , we invoke @xmath92 , the cyclicity of the trace and the product initial state ( [ eq - initial_state ] ) to get the matrix elements of @xmath93 in the joint eigenbasis @xmath94 of @xmath11 and @xmath12 , @xmath95 with @xmath96 evolving as for the free object , while the pointer matrix elements @xmath97 involve the bath evolution operator and shifted positions @xmath98 note that @xmath99 for @xmath100 . the hamiltonian @xmath13 can not be neglected in ( [ eq - rhos^0 ] ) , even for @xmath101 , although for such times @xmath102 . to evaluate the matrix elements ( [ eq - pointer_matrix_elements ] ) we use the high - temperature approximation @xmath103 for the pointer - bath gibbs state ( [ eqapp ] ) . given the weak - coupling condition @xmath104 implying the stability ( [ eq - stability ] ) ( thanks to @xmath105 , see @xcite ) , the error incurred is @xmath106 , as easily seen from the baker - campbell - haussdorf formula . thus @xmath107 where @xmath108 is the bath average with respect to the modified equilibrium state @xmath109 ; the normalization factor is determined by using the qclt as @xcite @xmath110 at this point we momentarily pause with dynamics and show that at @xmath23 , when @xmath111 , @xmath112 and @xmath113 , ( [ eq - y - integral ] ) yields the initial pointer state announced in ( [ eq - rhop(0 ) ] ) . to that end we invoke high temperatures @xmath114 again to approximate the matrix element @xmath115 by @xmath116 . replacing @xmath117 by @xmath118 in that expression and doing the gaussian @xmath119-integral in ( [ eq - y - integral ] ) , we arrive at the initial state ( [ eq - rhop(0 ) ] ) by neglecting terms @xmath120 . let us return to the time - evolved pointer matrix ( [ eq - y - integral ] ) . since @xmath121 factors into single - degree - of - freedom states , the qclt assigns gaussian statistics to the bath coupling agent @xmath36 for the average @xmath122 , with a mean @xmath123 given by linear response theory and a variance _ independent of y _ @xcite , @xmath124 therefore , @xmath125 coincides with its value for @xmath126 up to a phase factor @xmath127 and the @xmath119-integral in ( [ eq - y - integral ] ) remains gaussian for @xmath128 . of prime importance is the _ decoherence factor _ @xmath129 with a _ positive decoherence exponent _ @xmath130 revealed by the qclt as @xcite @xmath131 and a real phase @xmath132 irrelevant for decoherence . let us stress that the aforementioned results ( in particular ( [ eq - d_t ] ) ) are exact ( not lowest order in the pointer - bath coupling ) . they are consequences of wick s theorem as implied by the qclt and the additivity ( [ eq - int_hamiltonian ] ) of the bath coupling agent @xmath36 . direct proofs of ( [ eq - z_by ] ) , ( [ eq - varianceb ] ) and ( [ eq - d_t ] ) are easy in the particular case of a bath composed of harmonic oscillators linearly coupled to @xmath1 . it turns out @xcite that the phase factor @xmath133entails nothing but a correction of relative order @xmath134 to the decoherence exponent @xmath130 under the stability condition ( [ eq - stability ] ) . dropping that correction , the @xmath119-integral reduces to the initial pointer density matrix ( [ eq - rhop(0 ) ] ) , albeit with the shifted pointer positions @xmath135 and @xmath136 reflecting the action of the object - pointer coupling . our final result for the object - pointer state at time @xmath87 is @xmath137 with the notations specified in ( [ eq - rhop(0 ) ] ) , ( [ eq - rhos^0 ] ) and ( [ eq - d_t ] ) . entanglement and decoherence contribute separately in that remarkably simple `` final state '' ; they lead respectively to the second and third factors in ( [ eq - object - pointer - state ] ) . the decoherence ( third ) factor equals unity for @xmath138 and @xmath139 . given the narrow peaks ( of width @xmath140 ) of the initial pointer density matrix ( [ eq - rhop(0 ) ] ) at @xmath141 , one can appreciate the fate of the @xmath142 coherences in the final state ( [ eq - object - pointer - state ] ) by setting @xmath143 there . the decoherence factor then reads @xmath144 and reveals irreversible decay as soon as the time @xmath84 much exceeds the _ decoherence time _ @xmath145 ; we may define that time implicitly as @xmath146 . it can be shown @xcite that @xmath147 is an increasing convex function of time ( see the inset in figure [ fig - lett ] ) . the diagonal ( @xmath138 and @xmath139 ) terms in the final state ( [ eq - object - pointer - state ] ) give the probability density of pointer positions for fixed @xmath40 . that density has a sharp peak ( of width @xmath140 ) at @xmath148 . the peaks associated to distinct @xmath40 and @xmath149 begin to be resolved at the _ entanglement time _ @xmath150 . that time is related to @xmath151 by @xmath152 where @xmath153 is a dimensionless measure of the strength of the pointer - bath coupling . figure [ fig - lett ] shows @xmath5 as function of @xmath154 for three distinct choices of the bath correlator @xmath155 . decoherence time @xmath156 in units of @xmath6 against @xmath157 in a @xmath158-@xmath158 scale . we take @xmath159 ( kms relation ) and @xmath160 . the larger decay time of the bath correlator @xmath62 is then the thermal time @xmath161 . solid curves : @xmath162 ( from left to right ) . broken curves : approximate expressions ( [ t_dec_short_time]-[d_t - markov - ohmic ] ) for @xmath163 ( dashed lines ) and @xmath164 ( dotted lines ) . inset : decoherence exponent @xmath165 against @xmath166 ( @xmath167 ) . , width=264 ] let us note that , in analogy with the results of @xcite , the @xmath168 matrix elements of the reduced density matrix of @xmath0 , @xmath169 , decay to zero on a time scale @xmath170 much shorter than both @xmath171 and @xmath151 ( see ( [ eq - rhop(0 ) ] ) and ( [ eq - object - pointer - state ] ) ) . recalling that @xmath11 has a discrete spectrum , we denote by @xmath172 the minimum of @xmath173 over all pairs @xmath174 of eigenvalues present in the object initial state ( we suppose that @xmath175 if @xmath40 and @xmath149 belong to a part of the spectrum containing arbitrarily close eigenvalues , near an accumulation point , so that @xmath176 ) . at time @xmath177 , neighbouring peaks of the pointer densities can be resolved . each eigenvalue @xmath40 of the measured object observable is then uniquely tied up with `` its '' pointer position @xmath178 . if @xmath84 is also much larger than the maximum decoherence time @xmath179 ( @xmath84 being still smaller than @xmath9 and @xmath10 ) , the matrix elements ( [ eq - object - pointer - state ] ) for @xmath168 almost vanish for all values of @xmath180 . assuming moreover that the spectrum of @xmath11 is non - degenerate , object and pointer are in a separable mixed state , @xmath181 , with @xmath182 . hence , according to the born rule , with probability @xmath183 the object is in the eigenstate @xmath26 and the pointer is in a state @xmath184 localized near @xmath185 with probability density @xmath186 , in agreement with the von neumann postulate . the coupling @xmath38 may be switched off at time @xmath187 , where @xmath188 is defined in subsection 2.2 . then all pointer states @xmath189 are outside the effective potential well save for one eigenvalue @xmath190 . the inter - peak distance is amplified at time @xmath191 by the effective pointer dynamics , till it reaches a macroscopically resolvable magnitude @xmath82 . then a pointer reading , while still a physical process in principle perturbing @xmath1 , surely can not blur the distinction of the peaks . formula ( [ eq - dectime_vs_enttime ] ) explicitly yields the decoherence time in two interesting limits . an _ interaction dominated _ regime has decoherence outrunning bath correlation decay ( @xmath192 ) such that we can use @xmath193 in ( [ decexp ] ) and ( [ eq - dectime_vs_enttime ] ) . we conclude @xmath194 for @xmath195 . the decoherence time depends on the bath through the pointer - bath coupling strength @xmath196 only . it is smaller than @xmath8 when @xmath197 . the small - time behaviour ( [ t_dec_short_time ] ) of the decoherence factor appears also in other models @xcite . the opposite limit @xmath198 defines the _ markovian regime_. a rotating - wave approximation is inappropriate due to our restriction ( [ timescales ] ) . assuming that the fourier transform of the imaginary part of @xmath199 behaves as @xmath200 when @xmath201 , we find @xmath202 for @xmath203 . here @xmath204 for a `` ohmic bath '' ( @xmath205 ) and @xmath206 for a `` super - ohmic bath '' ( @xmath207 ) ; the constant @xmath208 is independent of the strengths of the couplings ( [ eq - int_hamiltonian ] ) , @xmath209 and @xmath210 the asymptotic results ( [ t_dec_short_time])-([d_t - markov - ohmic ] ) with numerical solutions of ( [ eq - dectime_vs_enttime ] ) . one finds an excellent agreement except for a thin range around @xmath211 . if @xmath212 , the only regime with a decoherence faster than resolution of pointer peaks ( @xmath213 ) is the markov regime ( @xmath214 ) with @xmath205 ( ohmic bath ) . in all asymptotic regimes @xmath215 if @xmath216 and @xmath217 . this means that mesoscopic superpositions decay to mixtures faster than entanglement can create them . we have investigated a model for quantum detection in which the entanglement produced by the coupling of the measured object with the pointer is simultaneous with decoherence of distinct pointer readouts ; the apparatus ( pointer and bath ) is taken initially in a local thermal metastable state , not correlated to the object . we have shown that the decoherence time @xmath5 presents a universal behaviour in the interaction - dominated regime @xmath218 , whereas it depends strongly on the small - frequency behaviour of the bath correlator in the markov regime @xmath219 . for reasonably strong pointer - bath and weak object - pointer couplings , @xmath5 is smaller than the time @xmath8 needed by entanglement to produce mesoscopic superpositions , which do not appear at any time during the measurement . several generalizations of our results present no difficulties . first , nonlinear pointer - bath couplings , @xmath220 with @xmath221 , make for richer decoherence scenarios and produce smaller decoherence times save for ohmic baths in the markovian regime @xcite . second , the qclt also works for baths of interacting particles if the correlator @xmath222 decays more rapidly than @xmath223 ( see @xcite for a related version of the qclt in this context ) . we shall publish these elaborations elsewhere . we acknowledge support by the deutsche forschungsgemeinschaft ( through the project sfb / tr 12 ) and the agence nationale de la recherche ( project anr-05-jcjc-0107 - 01 ) and thank m. guta for pointing out @xcite to us .	we study a class of quantum measurement models . a microscopic object is entangled with a macroscopic pointer such that a distinct pointer position is tied to each eigenvalue of the measured object observable . those different pointer positions mutually decohere under the influence of an environment . overcoming limitations of previous approaches we ( i ) cope with initial correlations between pointer and environment by considering them initially in a metastable local thermal equilibrium , ( ii ) allow for object - pointer entanglement and environment - induced decoherence of distinct pointer readouts to proceed simultaneously , such that mixtures of macroscopically distinct object - pointer product states arise without intervening macroscopic superpositions , and ( iii ) go beyond the markovian treatment of decoherence .
to date , more than 200 extrasolar planets have been discovered about the nearby stars within 200 pc ( butler et al . 2006 ; the extrasolar planets encyclopaedia ) mostly by the measurements of doppler surveys and transiting techniques . the increasing numbers of the extrasolar planets are greatly attributed to the increasing of precision of measurement . at present , the observational precision has been achieved to @xmath13 ms@xmath14 to @xmath15 ms@xmath14 , whereas most observations currently have lower precision . the improvement of the observations will indeed induce the substantial discovery : ( 1)plentiful multiple systems , which @xmath16 multiple system involved in orbital resonance and secular interactions now have been detected ; ( 2 ) much more low - mass companions around main - sequence stars ( from neptune - mass to earth - mass , so - called super - earths ) , e.g. , 55 cancri ( mcarthur et al . 2004 ) , gj 876 ( rivera et al . 2005 ) , hd 160691 ( santos et al . 2004 ; gozdziewski et al . 2006);(3)a true solar system analog , bearing several terrestrial planets , the asteroidal structure and the dynamical environment of potentially terrestrial planets in the habitable zones ( hzs ) ( kasting et al . 1993 ) for the development of life ; ( 4)the census of the systems , and the diversity of the planetary systems may provide abundant clues for theorists to more accurately model the planetary formation ( ida & lin 2004 ; boss 2006 ) . lovis et al . ( 2006 ) ( hereafter paper i ) recently reported the discovery of an interesting system of three neptune - mass planets orbiting about hd 69830 through high precision measurements with the harps spectrograph at la silla , chile . the nearby star hd 69830 is of spectral type k0v with an estimated mass of @xmath17 and a total luminosity of @xmath18 ( paper i ) , about 12.6 pc away from the sun . in addition , beichman et al . ( 2005 ) announced the detection of a large excess owing to hot grains of crystalline silicates orbiting the star hd 69830 and inferred that there could be probably a massive asteroid within 1 au . subsequently , alibert et al . ( 2006 ) and paper i performed lots of calculations to simulate the system and revealed that the innermost planet may possess a rocky core surrounded by a tiny gaseous envelope and probably formed inside the ice line in the beginning , whereas the two outer companions formed outside the ice line from a rocky embryo and then accreted the water and gas onto the envelope in the sequent formation process . hence , it is important for one to understand the dynamical structure in the final assemblage of the planetary system ( asghari et al . 2004 ; ji et al . 2005 ; jiang & yeh 2006 ) , and to investigate suitable hzs for life - bearing terrestrial planets ( david et al . 2003 ; fatuzzo et al . 2006 ; jones et al . 2005 ; haghighipour 2006 ) advancing the space missions ( such as _ corot , kepler _ and _ tpf _ ) aiming at detecting them , thus one of our goals is to focus on this issue ( 2.3 ) . on the other hand , from a dynamical viewpoint , one may be concerned about the long - term stability of the system . in the present work , we further perform a large number of simulations of this system to examine the stability ( 2.1 ) and to understand the secular evolution by means of the secular theory ( 2.2 ) . finally , we summary the main results and present a concise discussion in 3 . in this paper , we adopt the orbital parameters of the hd 69830 system provided by paper i. in the simulations , we always take the stellar mass and the minimum planetary masses from table 1 , except where noted . to be specific , the mass of the host star is @xmath19 , and the masses of three neptunian planets are respectively , @xmath20 , @xmath21 and @xmath22 , where @xmath23 . we utilize n - body codes ( ji , li & liu 2002 ) to perform numerical simulations by using rkf7(8 ) and symplectic integrators ( wisdom & holman 1991 ) for this system . in the numerical study , the adopted time stepsize is usually @xmath24 2% - 5% of the orbital period of the innermost planet . in addition , the numerical errors were effectively controlled over the integration timescale , and the total energy is generally conserved to @xmath25 for the integrations . the typical timescale of simulations of the three - planet system is from 100 myr to 2000 myr . to examine the secular stability of this system , we numerically integrated the three - planet system on a timescale of 2 gyr in terms of the initial values listed in table 1 . in figure 1 , a snapshot of the secular orbital evolution of three planets is illustrated , where @xmath26 , @xmath27 ( the subscipt @xmath28 , individually , denoting planet b , c and d ) are , respectively , the apoapsis and periapsis distances . in the secular dynamics , the semi - major axis @xmath29 remains unchanged to be 0.0785 au for 2 gyr , whilst @xmath30 and @xmath31 slightly undergo librations about 0.186 and 0.630 au with small amplitudes over the same timescale . the variations of eccentricities during long - term evolution are followed , where @xmath32 , @xmath33 and @xmath34 . in fig.1 , we note that the time behaviors of @xmath35 and @xmath36 again show regular motions of bounded orbits for three planets , indicating their orbits are well separated in the secular evolution due to smaller mutual interactions . in the numerical study , we find the system can be dynamically stable and last at least for 2 gyr , and our simulations do extend and confirm the numerical exploration of 1 gyr by paper i. to explore the stability of hd 69830 with respect to the variations of the planetary masses , we first fix @xmath37 in increment of 0.1 from 0.1 to 0.9 . in the additional numerical experiments , we simply alter the masses but keep all orbital parameters ( table 1 ) , again restart new runs of integration of three - planet system for 100 - 1000 myr with the rescaled masses . as a result , we find the system could remain definitely stable for the above investigated timescale with slight vibrations in semi - major axes and eccentricities for all planets . however , paper i also pointed out that the system can even survive for the planets bearing the jupiter - masses . to verify this , we further let @xmath38 , then the planetary masses are in the jupiter - mass range , and the outcomes for new integrations of 100 myr show that there is no chaotic behavior occurring in this system . this may also imply that the stability of the hd 69830 system is not so sensitive to the planetary masses . next , we come to the qualitative stability criteria for this system . gladman ( 1993 ) analytically attained the minimum separation @xmath39 between two planets of masses @xmath40 and @xmath41 moving on initially circular orbits required for stability , @xmath42 where the mutual hill radius @xmath43 is defined by @xmath44 with the orbital data in table 1 , we then approximately estimate @xmath45 au , for the couple of planet b and c ; similarly , @xmath46 au , for the pair of planet c and d. note this criteria are good guidelines for stability but are not definitive constraints for nearly - circular orbits . however , chambers et al . ( 1996 ) presented a more general stability criterion that shows the initial orbital separation should be no less than about 10 @xmath43 for multi - planet systems . for the neptune - mass planets , the actual orbital separations are @xmath47 times of the analytical values that the stability requires , while the mutual distances between two orbits can amount to be @xmath24 10 @xmath43 , even though these planets possess jupiter masses therefore , from both the numerical and analytical views , we can safely conclude that the hd 69830 system is dynamically stable for the lifetime of the star . considering a system of @xmath48 massive bodies of mass @xmath49 @xmath50 moving about the central star of mass @xmath51 under their mutual gravitational forces and the attraction of the central body , if @xmath52 , and these planets are with the orbits of small eccentricities and low inclinations , then the long - term evolution of the eccentricities and inclinations of all planets can be analytically described by classical laplace - lagrange ( hereafter l - l ) secular theory , provided that the mean motion resonance amongst them is absent . many authors have adopted the secular theory to study the orbital evolution of the planets in the systems , e.g. , ji et al . ( 2003 ) applied this theory to explore the dynamical evolution of three giant planets in 55 cancri ; more recently , adams & laughlin ( 2006 ; and references therein ) further investigated the dynamical interactions in the multiple planetary system by using secular theory and showed that the orbital eccentricities in these systems alter over secular time scales , of typical variations over @xmath53 years . as for the hd 69830 system , there are three neptune - mass planets orbiting the host star , assuming all planets are coplanar . furthermore , in table 1 we notice that the orbital eccentricities are smaller than about 0.10 , hence one may realize that the l - l secular theory can be applied to this system . as usual , the l - l secular theory tells us the story of the precession of the longitudes of periastron and that of the longitudes of node , and the variations of the amplitudes of the eccentricities and inclinations . herein , we simply address the issue of the secular evolution of the eccentricities . bearing this in mind , the l - l theory is an approximation to the actual motions , because it is valid only to the second order in the eccentricities and inclinations and the first order in the planetary masses , however , it may reveal substantial clues for long - term evolution of the system . subsequently , we will use the analytical results from l - l theory to compare with the numerical outcomes from direct integrations . the linear l - l secular solutions for the eccentricities ( the eccentricity eigenvectors , see murray & dermott 1999 , md99 ) can be written : @xmath54 @xmath55 where @xmath56 , @xmath57 and @xmath58 ( @xmath59 , respectively for companions b , c and d ; @xmath28 ) are respectively , the magnitudes of eccentricity eigenvectors , the eigenfrequencies and the phases , which can be calculated from the initial orbital solutions given in table 1 . in table 2 are listed the magnitudes of the eccentricity eigenvectors @xmath56 evaluated from the secular theory by considering all three planets . firstly , we obtain three eigenfrequencies @xmath60 yr@xmath14 , @xmath61 yr@xmath14 and @xmath62 yr@xmath14 ( see table 2 ) for the hd 69830 system , and these values are related to the precession of the periastron of three neptune - mass planets with periods of 5,491 yr , 16,248 yr and 152,650 yr , which are essentially in agreement with the results of paper i. in addition , the phases are @xmath63 , @xmath64 and @xmath65 , and now all constants in the right - side of eqs . ( [ eq1 ] ) and ( [ eq2 ] ) are well determined , hence we can derive the analytical expressions of the eccentricities @xmath66 for planet @xmath67 at any time @xmath68 , and its squares are given by @xmath69 secondly , we estimate the period of @xmath0 in the secular orbital evolution using the l - l theory , from table 2 , we note that the magnitudes of the eccentricity vectors @xmath70 , and it is not difficult for one to understand that the mutual interactions of two inner planets are much greater than that of the distant outermost planet . hence , we have @xmath71 then , the secular period of @xmath0 given by analytical method is @xmath72 , about 8,300 yr . similarly , the secular period of @xmath1 can be also derived , where @xmath73 , apparently , is equal to @xmath74 . now , we compare the analytical predictions with the numerical integrations . in fig . 2 , a snapshot is illustrated to perform the comparisons , while it can reflect the situation of total timescale of 2 gyr . in this figure , the _ dash dot line _ denotes the variational eccentricities of numerical results , while the _ solid line _ represents those from secular theory . for the upper and middle panels , @xmath74 and @xmath75 are clearly evident and also in accordance with the corresponding long - term periods for eccentricities in the numerical results . in the meantime , the librating amplitudes for the eccentricities @xmath0 , @xmath1 from the l - l theory ( _ solid line _ ) , where @xmath76 , then @xmath0 is in the range [ 0.05 , 0.20 ] and @xmath1 in [ 0.0 , 0.15 ] , are well coupled with those of the numerical results ( _ dash dot line _ ) by long - term integrations . moreover , we also notice that @xmath0 experiences to be maximum while @xmath1 goes down minimum in the orbital evolution ( see fig.2 ) , and _ vice versa _ , still this is because of the conversation of total angular momentum for the system . as aforementioned , the mutual perturbations between two inner planets are much more important than their interactions with the outer planet ( see table 2 ) . on the other hand , in the secular evolution over 2 gyr , @xmath29 and @xmath30 almost keep unchanged ( fig . 1 ) , by ignoring planet d , thus we have the total angular momentum @xmath77const ( @xmath78 are the combination constants of the planetary masses and semi - major axes ) , hence we obtain the maximum - minimum correlation of @xmath0 and @xmath1 , _ de facto _ , such orbital evolution may again imply two coupled secular periods of the eccentricity in this system . as to planet d , the contributions of the innermost planet ( hd 69830 b ) is much less than that of the intermediate companion ( hd 69830 c ) , indicating the mutual interaction from planet c and d is dominant , then we neglect the relevant eccentricity items arising from planet b in ( [ eq3 ] ) , and the secular period of the eccentricity of the outermost planet is achieved , where @xmath79 about 18,200 yr . for the bottom panel in fig . 2 , one can observe the long - term eccentricity periods of @xmath80 of the secular theory and numerical simulations coincide with each other . moreover , the secular theory presents the scope of @xmath4 in [ 0.069 , 0.078 ] , which also confirms the direct numerical simulations . lastly , we reach the conclusion that the outcomes from l - l theory are in accordance with the numerical results . modern observations by _ spitzer _ and _ hst _ clearly exhibit that the circumstellar debris disks ( e.g. , au mic and @xmath81 pic ) are quite common in the early planetary formation ( for a recent review see werner et al . 2006 ) , and beichman et al . ( 2006 ) show that there are @xmath82 for kuiper belt analogs around mature main sequence stars by _ spitzer _ programs , further they point out that the existence of debris disks is extreme important to the resulting detection of individual planets and related to the formation and evolution of planetary systems . as mentioned previously , beichman et al . ( 2005 ) also provide a clear evidence of the presence of the disk in hd 69830 , and subsequently paper i produce the best - fit orbital solutions of three neptune - mass planets with well - separated nearly - circular orbits , and meanwhile the present study confirms the secular orbital stability in this system , which may imply that the hd 69830 system is likely to be an analog of the solar system , then to possess other terrestrial planets , and to be made of a potential asteroidal belt structure or kuiper belt analog in the system . hence , it deserves to make a detailed investigation from a numerical perspective . to further investigate the dynamical structure and potential hzs in this system , we performed additional simulations with the planetary configuration of coplanar orbits of three neptune - mass companions with one massive earth - like planet . in this series of runs , the mass of the assumed terrestrial planet ranges from 0.01 @xmath83 to 1.0 @xmath83 . and the initial orbital parameters are as follows : the numerical investigations were carried out in [ @xmath84 parameter space by direct integrations , and for a uniform grid of 0.01 au in semi - major axis ( 0.07 au @xmath6 au ) and 0.01 in eccentricity ( @xmath85 ) , the inclinations are @xmath86 , and the remainder angles are randomly distributed between @xmath87 and @xmath88 for each orbit , then the low - mass bodies were numerically integrated with three neptune - mass planets in the hd 69830 system . in total , about 2400 simulations were exhaustively run for typical integration time spans from @xmath7 to @xmath89 yr ( about @xmath89 - @xmath90 times of the orbital period of the innermost planet ) . our main results now follow . in figure 3 , are shown the contours of the surviving time for earth - like planets ( fig . 3a ) and the status of their final eccentricities ( fig . 3b ) for the integration over @xmath7 yr , where the horizontal and vertical axes are the initial @xmath91 and @xmath92 . fig.3a displays that there are stable zones for the low - mass planets in the regime between 0.3 and 0.5 au , and 0.8 and 1.2 au with final eccentricities of @xmath8 . obviously , there exist unstable zones for the nearby orbits of three planets , where the planetary embryos there have short dynamical surviving time ( usually less than @xmath93 yr ) and the eccentricities can be quickly pumped up to a high value @xmath94 ( see fig . 3b ) , where the evolution is not so sensitive to the initial masses . in fact , these adjacent bodies are in association with many of the mean motion resonances of the neptunian planets and the overlapping resonance mechanism ( md99 ) can reveal their chaotic behaviors of being ejected from the system in short dynamical lifetime , furthermore most of their orbits are within @xmath95 sphere of the neptune - mass planets , and others are involved in the secular resonance with two inner companions . analogous to our solar system , if consider the middle planet c as the counterpart as jupiter , we will have the regions of mean motion resonances : 2:1 ( 0.117 au ) , 3:2 ( 0.142 au ) , 3:1 ( 0.089 au ) and 5:2 ( 0.101 au ) , 2:3 ( 0.244 au ) , and in fig . 3 we notice there indeed exist the apparent asteroidal gaps about or within above mmrs ( e.g. , 3:1 and 5:2 mmrs ) , while in the region between 0.10 au and 0.14 au for @xmath96 , there are stable islands that the planetary embryos can last at least @xmath7 yr . in addition , for planet d , most of the terrestrial planets in 0.50 au @xmath11 0.80 au are chaotic and their eccentricities are excited to moderate even high values , the characterized mmrs with respect to the accumulation or depletion of the asteroid belt are 3:2 ( 0.481 au ) , 2:1 ( 0.397 au ) , 5:2 ( 0.342 au ) , 4:3 ( 0.520 au ) , 1:1 ( 0.630 au ) , 2:3 ( 0.826 au ) , 1:2 ( 1.000 au ) , and our results enrich those of paper i for massless bodies over two consecutive 1000-year intervals , showing a broader stable region beyond 0.80 au . note that there exist stable trojan terrestrial bodies in a narrow stripe about 0.630 au , involved in 1:1 mmr with planet d , and they can survive at least @xmath89 yr with resulting small eccentricities in the extended integrations . the stable trojan configurations may possibly take place in the extrasolar planetary systems , for example , ji et al . ( 2005 ) explored such trojan planets moving about 47 uma , and gozdziewski & konacki ( 2006 ) also argued that there may exist trojan pairs configurations in the hd 128311 and hd 82943 systems . recently , ford & gaudi ( 2006 ) develop a novel method of detecting trojan companions to transiting close - in extrasolar planets and argue that the terrestrial - mass trojans may be detectable with present ground - based observatories , whereas the terrestrial trojan planets with low eccentricity orbits about 1 au would be probably a fascinating habitatable environment for other life - bearing worlds . malhotra ( 2004 ) and ji et al . ( 2005 ) showed that the secular resonances may possibly exist in the general planetary systems , and play a part in the dynamical distributions of small bodies , then shape the asteroidal architecture as well as the orbital resonances . in this system , if consider a planetary embryo with a mass of @xmath97 ( in lunar mass ) , the related location for @xmath9 secular resonance originating from planet b is @xmath98 au , where two eigenfrequencies for the terrestrial body and innermost planet are , respectively , @xmath99 yr@xmath14 and @xmath100 yr@xmath14 , which are obtained by the l - l secular theory , as the earth - like planet bears the finite mass and then it may modify the values of the _ in situ _ eigenfrequencies @xmath57 @xmath101 of three neptunian planets , thus @xmath102 , which also indicates that both planets almost share common secular apsidal precession rates in their motion . 3b exhibits that the eccentricity of the planetesimals could be excited by @xmath9 about 0.32 au , and meanwhile the 5:2 mmr about 0.34 au ( for planet d ) drives the eccentricities to modest values , which may imply that the combination effects of the secular resonance and mmr can act as a hybrid dynamical mechanism of clearing of the small bodies in the disk . similarly , @xmath10 resonance with respect to the intermediate companion ( planet c ) occurs for a lunar - size planet about @xmath103 au at an eigenfrequency of @xmath104 yr@xmath14 , and the @xmath10 region is also close to 1:2 resonance with the outermost planet d , where the eccentricities for low - mass bodies about 1 au can be stimulated to moderate values in the extended time integrations of @xmath89 yr . our results are in accordance with nagasawa et al . ( 2005 ) , who pointed out that the secular resonances can pass through the terrestrial region from outside to inside to excite the eccentricities of the separated planetary embryos . note that @xmath105 arises from planet d with an eigenfrequency of @xmath106 yr@xmath14 , which may be at work @xmath107 au . in a word , the numerical study may suggest that there exist asteroid structure of the hd 69830 system . in this work , we have studied the secular dynamics and dynamical structure of hd 69830 . we now summarize the main results as follows : ( 1)in the simulations , we show that the triplet neptunian system can be stable at least for 2 gyr and that the stability would not be greatly influenced when we shift the planetary masses from neptune - mass to jupiter - mass . the l - l theory can well describe and predict the secular orbital evolution for three neptune - mass planets , where both the secular periods and librating amplitudes in the eccentricities are well coincidence with those obtained from direct integrations . we first reveal that the secular periods of the eccentricity @xmath0 and @xmath1 are identical , about 8,300 yr . account for the nature of near - circular well - spaced orbits , the hd 69830 system may be a close analog of the solar system . nevertheless , as well - known in our solar system , jupiter can play a major role as a dynamical shield to bounce and throw away most of small bodies at times and further to prevent many asteroids from colliding into the earth ( or other inner terrestrial planets ) over the evolution . however , if all planets of hd 69830 are in neptune - mass , such protection mechanism for the potential terrestrial planets would be weaker and larger masses would be helpful in building a harmonic habitable environment . ( 2)moreover , we extensively investigate the planetary configuration of three neptune - mass companions with one massive terrestrial planet in the region of 0.07 au @xmath6 au to examine the existence of the potential earth - like planets and further to study the asteroid structure or hz in this system . we show that there are stable zones at least @xmath7 yr for the low - mass terrestrial planets locating between 0.3 and 0.5 au , and 0.8 and 1.2 au with final eccentricities of @xmath8 . furthermore , we also find that the secular resonance @xmath9 and @xmath10 arising from two inner planets can excite the eccentricities of the terrestrial bodies , and the accumulation or depletion of the asteroid belt are also shaped by orbital resonances of the outer planets , for example , the asteroidal gaps of 2:1 and 3:2 mmrs with planet c , and 5:2 and 1:2 resonances with planet d. on the other hand , the stellar luminosity of hd 69830 is lower than that of the sun , thus the hz should shift inwards compared to our solar system . however , kuchner ( 2003 ) pointed out that the habitable regime is considered to be centered at @xmath108 1.03 au @xmath109 , such that a water planet can exist in equilibrium with stellar radiation , where @xmath110 is the radius of the planet s orbit and @xmath111 is the luminosity of the host star , then @xmath110 is approximate to @xmath112 au for hd 69830 . therefore , in a dynamical consideration , the proper candidate regions for the existence of the potential terrestrial planets or hzs are 0.35 au @xmath11 0.50 au , and 0.80 au @xmath12 1.00 au for relatively low eccentricities . hopefully , the future space - based observations , e.g. , _ corot , kelper _ and _ tpf _ will definitely present a handful of samples belonging to the category of the terrestrial bodies . in final , we may summarize that the hd 69830 system can possess the asteroidal architecture resembling to the solar system and both the mean motion resonance ( mmr ) and secular resonances ( malhotra 2004 ; ji et al . 2005 ; nagasawa et al . 2005 ) will work together to influence on the distribution of the small bodies in the planetary system . however , the formation of this system is still a mystery and needs further investigation in forthcoming study . llll @xmath113sin@xmath114(@xmath83 ) & 10.2 & 11.8 & 18.1 + orbital period @xmath115(days ) & 8.667 & 31.56 & 197 + @xmath91(au ) & 0.0785 & 0.186 & 0.630 + eccentricity @xmath92 & 0.10 & 0.13 & 0.07 + longitude of peri . @xmath116(deg ) & 340 & 221 & 224 + epoch of peri.(jd ) & 2453496.8 & 2453469.6 & 2453358	we perform numerical simulations to study the secular orbital evolution and dynamical structure in the hd 69830 planetary system with the best - fit orbital solutions by lovis and coworkers ( 2006 ) . in the simulations , we show that the triplet neptunian system can be stable at least for 2 gyr and the stability would not be greatly influenced even if we vary the planetary masses from neptune - mass to jupiter - mass . in addition , we employ the laplace - lagrange secular theory to investigate the long - term behaviors of the system , and the outcomes demonstrate that this theory can well describe and predict the secular orbital evolution for three neptune - mass planets , where the secular periods and amplitudes in the eccentricities are in good agreement with those of the direct numerical integrations . we first reveal that the secular periods of the eccentricity @xmath0 and @xmath1 are identical about 8,300 yr , i.e. , @xmath2 ( @xmath3 are respectively , the eigenfrequencies of the system ) , while the secular variation of @xmath4 for the outermost planet has a period of @xmath5 yr . moreover , we extensively explore the planetary configuration of three neptune - mass companions with one massive terrestrial planet residing in 0.07 au @xmath6 au , to examine the asteroid structure in this system . we underline that there are stable zones at least @xmath7 yr for low - mass terrestrial planets locating between 0.3 and 0.5 au , and 0.8 and 1.2 au with final eccentricities of @xmath8 . still , we also find that the secular resonance @xmath9 and @xmath10 arising from two inner planets can excite the eccentricities of the terrestrial bodies , and the accumulation or depletion of the asteroid belt are also shaped by orbital resonances of the outer planets , for example , the asteroidal gaps at 2:1 and 3:2 mean motion resonances ( mmrs ) with planet c , and 5:2 and 1:2 mmrs with planet d. in a dynamical sense , the proper candidate regions for the existence of the potential terrestrial planets or habitable zones ( hzs ) are 0.35 au @xmath11 0.50 au , and 0.80 au @xmath12 1.00 au for relatively low eccentricities , which makes sense to have the possible asteroidal structure in the system .
the binary asymmetric channel @xmath5 is the discrete memoryless channel with binary alphabet and transition probability given by @xmath6 and @xmath7 , where @xmath8 denote the probability of receiving @xmath9 if @xmath10 was sent . without loss of generality we will assume @xmath11 and @xmath12 ( see discussion in section [ sectionfurtherremarks ] ) . + there are two extreme cases of binary asymmetric channels : the binary symmetric channels @xmath13 with @xmath14 and the z - channels @xmath15 with @xmath16 . we may say that all binary symmetric channels are equivalent from the point of view of encoding and decoding , and the same can be said regarding the z - channels . to be more precise , let @xmath17 and @xmath18 be two binary symmetric channels or two @xmath19-channels and let @xmath20 denotes the error probability of the code @xmath21 according to the channel @xmath22 . on the encoding part , given two @xmath0-block codes @xmath23 ( for some @xmath24 ) we have that @xmath25 if and only if @xmath26 . this means that if @xmath27 has smaller error probability than @xmath28 with respect to a specific bsc ( z - channel ) the same happens for all bscs ( z - channels ) . on the decoding part , for every @xmath0-block code @xmath21 we have @xmath29 for all @xmath30 . this means that the maximum likelihood decoding does not depend on the bsc ( or z - channel ) considered . these characteristics of the extreme channels does not hold for general binary asymmetric channels . that is , on the encoding part , there are channels @xmath17 and @xmath18 neither of them binary symmetric channel nor z - channel and codes @xmath31 verifying @xmath32 and @xmath33 . on the decoding part , for every fixed @xmath24 , there exist channels @xmath17 and @xmath18 neither of them binary symmetric channel nor z - channel and a code @xmath34 such that @xmath35 for some @xmath36 , consequently these channels determine different mld criteria . in this paper we focus on the problem of determining when two binary asymmetric channels determine the same mld criteria , when restricted to @xmath0-block codes . a simple necessary and sufficient condition for two channels @xmath5 and @xmath37 to determine the same decoding criterion is given . this condition ( which depends on @xmath0 ) induces an equivalence relation on the parameter space and we obtain an explicit formula for the number of non - equivalent channels . + we remark that the interest in binary asymmetric channels has increased due to applications in flash memories @xcite and as models in other areas , such as neuroscience @xcite . most of the work developed on binary asymmetric channels has focused on coding properties @xcite and on the design of codes with given properties , either for general asymmetric channels @xcite or for the @xmath19-channel @xcite . the substantial efforts devoted to the encoding perspective contrasts with the situation concerning the decoding aspect , the focus of this work . + this paper is organized as follows : in section [ sectionmldequivalence ] we introduce an equivalence relation between channels in such a way that equivalent channels determine equal decoding criteria when mld is considered and discuss some properties of this relation . we consider for each ( fixed ) @xmath38 the above equivalence relation restricted to the parametric family of channels @xmath39 . a description of the regions in the parameter space determined by this equivalence relation and the number of such regions is provided , for each @xmath38 , in section [ sectiondecisioncriteria ] . in section [ sectionfurtherremarks ] we discuss some properties of the bac - function and the areas of the regions determined by its level curves , which are related to the probability of a given channel to have some prescribed admissible criterion . we start by introducing an equivalence relation ( depending on @xmath40 ) that characterizes when two channel determine the same decoding criteria if mld is considered over all @xmath0-block codes . + let @xmath22 be a discrete memoryless channel with input and output alphabet @xmath41 and @xmath40 . for @xmath42 we denote by @xmath43 the probability of receiving @xmath9 if @xmath10 was sent through the channel @xmath22 . when @xmath44 we denote this probability by @xmath45 and if the channel is understood we denote it simply @xmath8 . let @xmath46 , for @xmath47 , be two discrete memoryless channels with input and output alphabet @xmath41 and @xmath40 . we say that these channels are @xmath0-equivalent ( denoted by @xmath48 ) if for every @xmath0-block code @xmath49 and every word @xmath50 , we have @xmath51 where we interpret @xmath52 as returning lists of size at least @xmath53 . we remark that two channels @xmath17 and @xmath18 are @xmath0-equivalent if and only if the maximum likelihood decoding on @xmath17 coincides with the maximum likelihood decoding on @xmath18 when restricted to @xmath0-block codes . + let @xmath54 be a discrete memoryless channel . the @xmath0-fold associated channel is a channel @xmath55 whose transition probability is given by @xmath56 for @xmath57 . in particular the @xmath0-fold channel associated with @xmath44 is denoted by @xmath58 . we identify the set @xmath59 with the elements of @xmath60 via its binary expansion . under this identification , we associate with the @xmath0-fold binary asymmetric channel @xmath61 a matrix @xmath62 whose @xmath63-entry is equal to @xmath64 where @xmath65 and @xmath66 is the @xmath67-th bit of @xmath9 and @xmath10 respectively . for example , the matrix @xmath68 is a @xmath69 matrix whose entry in the row @xmath70 and column @xmath71 is @xmath72 . + with this terminology , we remark that two channels @xmath17 and @xmath18 are @xmath0-equivalent if and only if their associated @xmath0-fold channels @xmath73 and @xmath74 are @xmath53-equivalent . when two channels are @xmath53-equivalent we say that their are mld - equivalent . + when considering mld - equivalence , we need to compare the entries in each line of the channel matrix and it is useful to replace it by a simpler matrix . in @xcite , the authors substituted a matrix @xmath75 by a matrix @xmath76 with entries @xmath77 if @xmath78 is the @xmath79-th largest element ( allowing ties ) of the @xmath80-th column . in order to determine whether two channels are equivalent we adopt a different , but similar substitution . let @xmath75 be an @xmath81 real matrix . the _ ordered form _ of @xmath75 is the integer matrix @xmath82 defined by @xmath83 . the ordered form @xmath84^{}$ ] of the channel matrix @xmath84 $ ] will be used to characterize the equivalent @xmath0-fold bacs . we observe that @xmath85 for all @xmath86 . the following equivalences are immediate consequence of the definition of @xmath84^{}$ ] . let @xmath87 be two discrete channels with input and output alphabet @xmath88 . let @xmath89 and @xmath90 denote the matrices of the @xmath0-fold channels @xmath73 and @xmath74 respectively . the following assertions are equivalent . 1 . the channels @xmath17 and @xmath18 are @xmath0-equivalent . @xmath91 for all @xmath92 . 3 . @xmath93 . let @xmath94 be two discrete channels and @xmath95 . considering the map @xmath96 given by @xmath97 , it is easy to check that @xmath98 implies @xmath48 . when @xmath48 for all @xmath24 we say that these channels are @xmath99-equivalent ( in this case we denote @xmath100 ) . + we consider the parameter space @xmath101 for the binary asymmetric channel and let @xmath24 be a fixed integer or @xmath102 . the @xmath0-equivalence for bacs induces an equivalence relation in @xmath103 : @xmath104 if and only if @xmath105 . we denote by @xmath106 , the set of equivalence classes and by @xmath107 the projection that associates @xmath3 to its equivalence class ( i.e. @xmath108 ) . the main result of this paper is a complete description of the quotient set @xmath109 . a decision criterion of order @xmath0 for the bac is an equivalence class @xmath110 ( in particular @xmath111 is a subset of @xmath103 ) . an @xmath0-_stable _ decision criterion @xmath111 is a decision criterion of order @xmath0 for the bac which is an open set of @xmath103 and an @xmath0-_unstable _ decision criterion is a decision criterion of order @xmath0 for bac with no interior points . if @xmath111 is a stable decision criterion for the @xmath0-fold bac and @xmath112 then maximum likelihood decoding on @xmath58 remains the same under small perturbation of the parameter @xmath3 . a point @xmath113 is @xmath0-_stable _ if @xmath3 is an interior point of @xmath114 and @xmath0-_unstable _ otherwise . the @xmath0-stable region ( denoted by @xmath115 ) is the set of all @xmath0-stable points and the @xmath0-unstable region ( denoted by @xmath116 ) is the set of all @xmath0-unstable points . we will prove later that every decision criterion is either stable or unstable , that is , if a criterion contain an @xmath0-stable point then all its points are @xmath0-stable . we conclude this section discussing how the parameter space @xmath103 decomposes into different @xmath0-equivalence classes for @xmath117 . + for @xmath118 we have @xmath119 and the matrix of @xmath120 is given by @xmath121 , thus its ordered form @xmath122 does not depend on @xmath3 and we have only one criterion , which is stable . + for @xmath123 we have @xmath124 and the matrix @xmath125 of @xmath126 ( with @xmath113 ) is given by @xmath127 in this case the ordered form depends on @xmath3 : * if @xmath3 is an interior point of @xmath103 the ordered form is given by @xmath128 ; * if @xmath129 ( and hence @xmath130 , since @xmath131 ) we obtain the ordered form @xmath132 ; * if @xmath133 we obtain @xmath134 . in this case we have three different decision criteria , one stable and two unstable . + for @xmath135 and @xmath136 we started with some simulations using the software sage @xcite . we considered a set @xmath137 of @xmath138 points uniformly distributed on @xmath103 and calculated the ordered form @xmath139 of each @xmath140 . + for @xmath141 we observe two criteria , @xmath142 and @xmath143 , and by coloring the points in these regions by blue and red respectively we obtain the picture showing in figure [ triangulon3 ] . -fold bac . ] as we will prove later ( theorem [ maintheorem ] ) , there are two stable criteria @xmath142 and @xmath143 ( the connected components of the stable region @xmath144 ) and there are three unstable decision criteria corresponding to the curves @xmath129 ( the @xmath19-channel ) , @xmath133 ( the bsc ) and the curve that separates the two connected components of the stable region . + we proceed similarly with the cases @xmath145 and @xmath146 , observing three decision criteria for @xmath145 ( figure [ triangulon4 ] ) and five decision criteria for @xmath146 ( figure [ triangulon5 ] ) . they correspond to the stable decision criteria and the curves separating these regions correspond to the four and six different unstable decision criteria for @xmath145 and @xmath146 respectively ( see theorem [ maintheorem ] in the next section ) . -fold bac . ] -fold bac . we start introducing a function which plays a fundamental role in describing the regions which determine the decision criteria for the bac . this function also induces a natural distance between binary asymmetric channels ( see section [ sectionfurtherremarks ] for more details ) . let @xmath147 $ ] be the function given by @xmath148 if @xmath149 , and @xmath150 if @xmath129 . we refer to this function as the bac - function . it is easy to check that in fact the image of @xmath151 is contained in the interval @xmath152 $ ] where the values @xmath153 and @xmath53 are attained by the extremes cases @xmath129 and @xmath133 , respectively . besides , this function is continuous in the connected set @xmath103 and therefore we have @xmath154 $ ] . + let @xmath155 and @xmath156 be integers with @xmath157 . for @xmath113 we have @xmath158 where equality corresponds to equality . the next lemma is a direct consequence of the above relation . [ corimportante ] let @xmath155 and @xmath156 be natural numbers with @xmath159 , @xmath160 and @xmath161 . consider the words @xmath162 . then @xmath163 if and only if @xmath164 , where equality corresponds to equality . in particular , if @xmath165 and @xmath166 the channels @xmath167 and @xmath168 are not @xmath0-equivalent . the above lemma establishes a condition for the @xmath0-fold channels @xmath169 and @xmath170 not being mld - equivalent . when we consider @xmath171 and @xmath172 as variables , the entries of the transition matrix @xmath173 of the channel @xmath174 are polynomials in the variables @xmath171 and @xmath172 . in fact , the entry of @xmath173 corresponding to the conditional probability @xmath8 ( @xmath175 ) is equal to the polynomial @xmath176 where @xmath177 and @xmath178 correspond to the number of indices @xmath67 for which @xmath179 is equal to @xmath180 and @xmath181 , respectively . in particular we have @xmath182 and the hamming weight of @xmath9 is equal to @xmath183 , which we will refer also as the weight of @xmath184 and denote by @xmath185 . we remark that two polynomials in the same row of @xmath173 have necessarily the same weight . by the previous consideration we define the following sets of bivariate polynomials : * @xmath186 . * @xmath187 for @xmath188 . to determine the decision criterion of an @xmath0-fold bac we need to do comparisons only between values in the same row of its transition matrix what means comparisons of values of polynomials belonging to @xmath189 for some @xmath79 , @xmath188 . the next lemma describes the stable region in terms of these sets . [ lemmaregstable ] the @xmath0-stable region @xmath115 is given by @xmath190 we denote by @xmath191 the set on the right side of the above equality . this set is open in @xmath103 ( since it is a finite intersection of open sets in @xmath103 ) , therefore if @xmath192 there is a ball @xmath193 with center at this point such that @xmath194 . since @xmath195 is connected , the signal of @xmath196 does not depends on @xmath197 ( whenever @xmath198 for some @xmath79 ) and the same occurs with their associated decision criteria , so @xmath199 . to prove the other inclusion we suppose by contradiction that there exists @xmath200 verifying @xmath201 . then , there exists two distinct polynomials @xmath198 for some @xmath202 and a ball @xmath193 centered at @xmath203 such that @xmath204 and every point in @xmath195 determines the same decision criterion , in particular @xmath205 for all @xmath206 . since @xmath207 has interior points and two polynomial that coincide in an open set must be equal , we have @xmath208 which is a contradiction . as we will prove next , the main property of the bac - function from the point of view of this work , is that the curves which separate the regions corresponding to the stable and unstable criteria are level curves of this function associated with rational values . first , we prove some lemmas . [ lemmaregunstable ] let @xmath113 be an @xmath0-unstable point for the bac . then @xmath209 where @xmath210 are integers verifying @xmath211 and @xmath160 . by lemma [ lemmaregstable ] , if @xmath212 is an @xmath0-unstable point for the bac then there exists distinct polynomials @xmath213 for some @xmath214 such that @xmath215 . we write @xmath216 with @xmath217 and @xmath218 for @xmath47 . without loss of generality we suppose @xmath219 . let @xmath220 and @xmath221 , then we have @xmath222 or , equivalently , @xmath223 evaluating the equation [ eqlevelcurves ] for @xmath224 and using the relation ( [ lemmaeqforlevelcurves ] ) we have @xmath225 . by our assumption we have @xmath226 , since @xmath154 $ ] then @xmath227 and @xmath228 ; simplifying common factors if necessary we can assume @xmath229 . based on the above result , we introduce the following definition . the weight of a non - negative rational @xmath230 ( denoted by @xmath231 ) , is the sum of its numerator and denominator in the reduced expression of @xmath230 . an @xmath0-critical value for the bac - function @xmath151 ( where @xmath38 ) is a rational number @xmath232 $ ] with @xmath233 . the set of all @xmath0-critical values for @xmath151 is denoted by @xmath234 . if we write the set of the @xmath0-critical values for @xmath151 as @xmath235 and denote by @xmath236 then @xmath237 for @xmath238 . [ lemmacontinuity ] let @xmath113 and @xmath173 be the transition matrix for the channel @xmath39 ( seeing as an element of @xmath239 ) . the function @xmath240 given by @xmath241 is continuous . let @xmath242 and @xmath198 for some @xmath79 , @xmath188 with @xmath243 . by lemma [ lemmaregstable ] we have @xmath244 , then there exists @xmath245 such that the sign of @xmath196 does not depend on @xmath246 . if @xmath247 and denoting by @xmath248 the @xmath249-ball centered at @xmath203 we have that @xmath250 for all @xmath251 and for all @xmath252 . therefore @xmath253 , so @xmath254 is locally constant and in particular continuous . [ lemmaconexity ] let @xmath151 be the bac - function . then @xmath255 is a connected set for every interval @xmath256 $ ] . let @xmath257 and @xmath258 \rightarrow [ 0,1]$ ] be the function given by @xmath259 . we affirm that it is increasing ( in the variable @xmath171 ) . to prove this we consider @xmath260 $ ] and the function @xmath261 ( where @xmath262 ) . since @xmath263 we have @xmath264 and @xmath265 , therefore @xmath266 @xmath267 since @xmath268 and @xmath269 ( because @xmath270 and @xmath271 ) , then there is an unique @xmath272 such that @xmath273 , or equivalently , such that @xmath274 . therefore @xmath258 \rightarrow [ 0,1]$ ] is increasing since it is a continuous bijection with @xmath275 and @xmath276 . let @xmath277 $ ] be an interval and @xmath278 be two points in @xmath255 . we denote by @xmath279 and @xmath280 for @xmath281 . without loss of generality we assume @xmath282 . since @xmath283 there exists @xmath284 such that @xmath285 and @xmath286 for all @xmath287 ( in particular @xmath288 for all @xmath289 $ ] ) . taking the path @xmath290\rightarrow \mathcal{t}$ ] given by @xmath291 and @xmath292 the segment of curve in @xmath293 from @xmath294 to @xmath295 , we have that the concatenation path @xmath296 is a path connecting @xmath203 with @xmath295 . therefore @xmath255 is path - connected and then connected . [ lemma6 ] let @xmath203 and @xmath295 be two points in @xmath103 . if @xmath297 then @xmath298 for all @xmath299 . we suppose by contradiction that @xmath203 and @xmath295 are not @xmath0-equivalent , so there exists two polynomials @xmath198 for some @xmath79 verifying @xmath300 and @xmath301 . as in the proof of lemma [ lemmaregunstable ] , the polynomials @xmath184 and @xmath302 must verify an equation similar to equation ( [ eqlevelcurves ] ) : @xmath303 for some integers @xmath304 and @xmath227 satisfying @xmath159 and @xmath160 . we consider a path @xmath305 contained in the curve @xmath306 connecting @xmath203 with @xmath295 , since @xmath307 in @xmath203 and @xmath308 in @xmath295 , by continuity there exists an intermediate point @xmath309 in @xmath305 for which @xmath310 . evaluating equation ( [ eqlevelcurves2 ] ) in @xmath311 and using the relation ( [ lemmaeqforlevelcurves ] ) we have @xmath312 . since @xmath309 belongs to @xmath305 which is contained in the level curve @xmath306 we have @xmath313 , then @xmath314 . substituting @xmath224 in equation ( [ eqlevelcurves2 ] ) and using the relation [ lemmaeqforlevelcurves ] we have @xmath315 which is a contradiction . therefore @xmath203 and @xmath295 must be @xmath0-equivalent . now we are ready to state the main result of this paper . [ maintheorem ] let @xmath316 $ ] be the bac - function @xmath317 for @xmath149 , @xmath318 and @xmath319 its set of @xmath0-critical values ( @xmath38 ) . we consider the level curves @xmath320 for @xmath321 and the regions @xmath236 for @xmath322 . then @xmath323 where @xmath324 denotes the euler s totient function and there are exactly @xmath325 stable decision criteria for the @xmath0-fold bac which are given by @xmath326 and exactly @xmath327 unstable decision criteria for the @xmath0-fold bac given by @xmath328 . consider @xmath103 written as the disjoint union @xmath329 we have to prove that each @xmath330 and each @xmath331 is a decision criterion ( i.e. an equivalent class in @xmath109 ) . if @xmath332 for some @xmath333 , by lemma [ lemma6 ] and lemma [ corimportante ] we have @xmath334 if and only if @xmath335 , then each @xmath331 is a decision criterion . we consider now a point @xmath336 for some @xmath337 and @xmath338 . if @xmath339 for some @xmath80 , since each @xmath340 is a decision criterion we have @xmath341 . otherwise @xmath342 for some @xmath343 and there exists @xmath344 such that @xmath345 or @xmath346 . in both cases , by lemma [ corimportante ] we obtain @xmath341 . it only remains to prove that if @xmath347 the channels @xmath348 and @xmath349 are mld - equivalent . we consider the function @xmath240 given by @xmath241 where @xmath173 denotes the transition matrix for the channel @xmath39 . by lemma [ lemmaconexity ] and lemma [ lemmaregunstable ] then @xmath330 is a connected set contained in the stable region @xmath115 . by lemma [ lemmacontinuity ] the set @xmath350 is connected and since @xmath351 is discrete , there exists @xmath352 such that @xmath353 . therefore @xmath354 and @xmath334 . + since the @xmath330 are open sets for @xmath355 and the sets @xmath331 have empty interior they correspond to the stable and unstable criteria respectively . to derive the formula for @xmath325 we consider the decomposition into disjoint sets : @xmath356 where @xmath357 . we have @xmath358 for @xmath359 and for @xmath360 if @xmath361 then @xmath362 , in this case : @xmath363 @xmath364 @xmath365 then , for @xmath366 we have : @xmath367 for @xmath123 we have @xmath368 and the above formula also holds in this case . let @xmath278 . the channels @xmath167 and @xmath168 are @xmath99-equivalent if and only if @xmath297 if @xmath369 there exists a rational number @xmath370 for some @xmath371 large enough such that @xmath372 , then the channels @xmath167 and @xmath168 can not be @xmath373-equivalent . the converse is consequence of lemma [ lemma6 ] . a point @xmath113 is a stable point for the @xmath0-fold channel for all @xmath24 if and only if @xmath374 is irrational . using the formula for the average order for the euler s totient function @xmath324 ( see for example theorem 3.7 of @xcite ) we obtain the following corollary . the number of stable decision criteria for the @xmath0-fold bac grows quadratically with @xmath0 . more explicitly , it is given by @xmath375 . for @xmath146 ( see figure [ triangulon5 ] ) , we have the set of critical values @xmath376 which correspond to the level curves of the bac function describing the unstable sets . those curves can be seen from left to right according the order in @xmath377 . the five stable regions are the ones bounded by these curves . for @xmath378 , we have @xmath379 decision regions , @xmath380 instable , associated to the critical set @xmath381 @xmath382 and @xmath383 stable , situated between the level curves of the bac function attached to values in @xmath384 . to study the different criteria for the @xmath0-fold bac , we choose the parameter space @xmath385 : p+q<1 , 0\leq p \leq q\}\setminus \{(0,0)\}$ ] and use the bac - function to describe the regions determined by these criteria . in the first part of this section we discuss what happens when we remove the restriction @xmath12 and @xmath386 and what is the role of the bac - function in these cases . we also show how to obtain a natural distance between @xmath0-fold bacs in such a way that the bac - function measures how far a channel is from the binary symmetric channel , in this sense the bac - function can be interpreted as a measure of the asymmetry of the channel . + we remark first that condition @xmath12 is equivalent to the channel @xmath387 being reasonable ( in the sense of @xcite ) . more explicitly , for the channel @xmath39 we have the following facts ( which are consequence of the equation ( [ eqlevelcurves ] ) ) : * @xmath388 for all @xmath389 , * @xmath390 for all @xmath389 , * @xmath391 for all @xmath389 . as a consequence , we have that if two channels @xmath39 and @xmath392 are mld - equivalent then the sign of @xmath393 and @xmath394 is the same . the case @xmath395 corresponds to the completely noisy channel . these channels are all mld - equivalent among them and are not interesting from the coding / decoding viewpoint . the case @xmath12 can be decomposed into two regions @xmath103 and @xmath396^{2 } : p+q<1,\ 0\leq q \leq p\}\setminus \{(0,0)\}$ ] which are symmetric one to the other ( via the map @xmath397 ) . we observe that the bac - function @xmath151 can be extended to a function @xmath398 $ ] defining @xmath399 for @xmath400 if @xmath401 and @xmath402 . this extension is continuous and verifies @xmath403 $ ] . by relation ( [ lemmaeqforlevelcurves ] ) , @xmath404 if and only if @xmath405 if and only if @xmath406 , therefore for @xmath407 , @xmath408 and @xmath409 we have : * @xmath410 if @xmath113 with @xmath411 , * @xmath412 if @xmath413 with @xmath414 and * @xmath412 if @xmath133 . thus , we conclude that the triangles @xmath103 and @xmath415 have no common criteria decision except for those points corresponding to bsc . denoting by @xmath416 the matrix of @xmath39 , since the @xmath67-th row of @xmath416 is just the @xmath417-th row of @xmath418 in reverse order , then @xmath419 if and only if @xmath420 for all @xmath421 . by the above consideration we have the following proposition . let @xmath422 $ ] be the bac - function defined as above and @xmath319 its set of @xmath0-critical values ( @xmath38 ) where @xmath423 . there are exactly @xmath424 stable decision criteria for the @xmath387 with @xmath425 and exactly @xmath426 unstable decision criteria . the stable criteria are given by @xmath236 and @xmath427 for @xmath322 . the unstable criteria are given by the level curves @xmath428 and @xmath429 for @xmath321 . the function @xmath151 is constant when restricted to a criterion @xmath430 , thus it defines a injective function @xmath431 $ ] such that @xmath432 for any @xmath433 . if @xmath434 denotes the set of all criteria for the @xmath0-fold bac except for those corresponding to the @xmath19-channels ( i.e. when @xmath435 ) , then the function @xmath436 given by @xmath437 defines a metric in the criteria space for the @xmath0-fold bac channel . in particular if @xmath438 denotes the criterion corresponding to the bsc then @xmath439 can be interpreted as a measure of how asymmetric a channel is . + since the ordered form of the matrix associated with the completely noisy channels ( @xmath395 ) is the null matrix ( because in this case @xmath440 depends only on the hamming weight of @xmath9 and not on @xmath10 ) and this is the only situation for which it happens , then the points in the line @xmath395 correspond to a single criterion when considering @xmath0-fold bac s with @xmath441^{2}$ ] . we also saw that points with @xmath12 and @xmath442 can not be equivalent ( since the first corresponds to reasonable channels and the last not ) . let @xmath443 ^ 2 : p+q<1\}\setminus \{(0,0)\}$ ] and @xmath444 ^ 2:p+q>1)\}\setminus\{(1,1)\}$ ] . the involution @xmath445 maps the triangle @xmath446 into @xmath447 and the curves @xmath448 into itself . moreover , the @xmath67-th row of @xmath173 and @xmath449 are the same but in reverse order . therefore for @xmath450 , the channels @xmath39 and @xmath392 are mld - equivalent if and only if @xmath451 and @xmath452 are mld - equivalent . we conclude that if we consider @xmath0-fold bac s with parameters @xmath441^{2}$ ] , the stable criteria are the regions bounded by the edges of the square @xmath152 ^ 2 $ ] and the curves @xmath453 with @xmath454 positive rational number with @xmath160 ( @xmath455 ) . we also remark that the level curves @xmath453 contain the line @xmath456 but if we divide the equation that defines it by @xmath456 we obtain irreducible curves @xmath457 ( see figure [ square ] ) . -fold bac with @xmath3 in the unit square : the line @xmath395 ( dotted ) and the curves @xmath458 and @xmath459 for @xmath460 ( brown ) , @xmath461 ( violet ) , @xmath462 ( blue ) , @xmath463 ( orange ) , @xmath464 ( sky blue ) , @xmath465 ( green ) , @xmath466 ( red ) , @xmath467 ( yellow ) and @xmath53 ( black ) . curves corresponding to reciprocal values have the same color . ] by the previous discussion , without loss of generality we can restrict our parameter space to @xmath385 ^ 2 : p+q<1,\ p\leq q\}\setminus \{(0,0)\}$ ] . let @xmath468 be a criterion for the @xmath0-fold bac . if we choose a point @xmath113 uniformly at random and consider the channel @xmath469 , the probability @xmath470 where @xmath471 is the area of the region corresponding to the criterion @xmath472 . in this sense , the area of a given criterion for the @xmath0-bac is a measure of how probably this criterion is to be chosen ( assuming uniform distribution ) . by theorem [ maintheorem ] , @xmath473 where @xmath474 and @xmath475 are consecutive rational numbers in @xmath476 and @xmath477 is the area of the region @xmath478 . this area is equal to @xmath479 let @xmath314 where @xmath155 and @xmath156 and coprime positive integers with @xmath362 . applying the change of variable formula for double integrals with @xmath480 and after some calculations we obtain : @xmath481 since @xmath482 is a zero of order @xmath483 of the numerator , the integral is a proper integral . in some cases , a primitive for the integrand can be calculated explicitly , for example when @xmath484 and @xmath485 obtaining @xmath486 and @xmath487 . in the other cases we have used the software wolfram mathematica @xcite to calculate the integral ( [ areaequation ] ) numerically after some reductions . + in particular , the stable criterion for the @xmath0-fold bac nearest the @xmath0-fold bsc ( denoted by @xmath488 ) is given by the criterion corresponding to the channels @xmath39 for @xmath489 ( which we refer as @xmath0-fold quasi - symmetric channels ) . clearly @xmath490 when @xmath491 ( since @xmath492 ) . in the next table we show the percentages ( rounded to the nearest integer ) represented by the different stable criteria for the @xmath0-fold bac , @xmath493 . c\|c\|c\|c\|c\|c\|c\|c\|c\|c\|c + @xmath141 & & & + @xmath145 & & & & + @xmath146 & & @xmath494 & & @xmath495 & & + @xmath496 & & @xmath497 & @xmath494 & & @xmath495 & & + @xmath498 & @xmath499 & @xmath500 & @xmath497 & @xmath494 & @xmath501 & @xmath499 & @xmath495 & @xmath501 & @xmath502 & + for example , for @xmath141 the region corresponding to the channels @xmath503 with @xmath504 ( the @xmath505-fold quasi - symmetric channels ) represents the @xmath506 of the total area . the last percentage in each row of this table corresponds to the criterion @xmath488 , associated with the @xmath0-fold quasi - symmetric channels . by ( [ equationtn ] ) , the number of stable regions for @xmath507 and @xmath378 is @xmath508 and @xmath509 respectively . the percentages represented by these regions ( from left to right ) are @xmath510 @xmath511 for @xmath507 and @xmath512 @xmath513 , for @xmath378 . as we can see , the criterion @xmath488 maximizes the probability among all the criteria for the @xmath0-fold channels for @xmath514 . we have checked also for @xmath515 ( see figure [ dim40 ] to see graphically the percentages represented by all regions for @xmath516 ) and conjecture that this is true for any @xmath517 . let @xmath518 be the stable criterion for the @xmath0-fold bac nearest to the @xmath0-fold @xmath19-channel . we also point out an interesting comparison regarding the sizes of the areas corresponding to the criteria @xmath488 and @xmath518 . by considering for each @xmath0 the ratios @xmath519 and @xmath520 between the areas corresponding to @xmath521 and @xmath518 and the average area for this @xmath0 , it can be observed from our data that @xmath519 grows ( with some very small oscillation ) with @xmath0 linearly ( i.e. @xmath522 approaches @xmath79 ) whereas @xmath520 gets near to one . as a sample , for @xmath523 and @xmath524 , the obtained sequences @xmath525 are @xmath526 @xmath527 @xmath528 @xmath529 + @xmath530 and @xmath531 . stable regions for @xmath516 ordered by its bac - function values . the rightmost bar corresponds to @xmath532 . ] work partially supported by cnpq grants 150270/2016 - 0 , 140368/2015 - 9 , + 303985/2014 - 3 and 312926/2013 - 8 and by fapesp grants 2015/26420 - 1 + and 2013/25977 - 7 . y. cassuto , m. schwartz , v. bohossian , j. bruck , `` codes for asymmetric limited - magnitude errors with application to multi - level flash memories , '' ieee transaction on information theory , vol . 56 , no . 4 , pp . 1582 - 1595 , 2010 .	this work concerns with the @xmath0-fold binary asymmetric channels ( @xmath1 ) . an equivalence relation between two channels can be characterized by both having the same decision criterion when maximum likelihood is considered . we introduce here a function @xmath2 ( the bac - function ) such that the parameters @xmath3 of the binary channel which determine equivalent channels belong to certain region delimited by its level curves . explicit equations determining these regions are given and the number of different @xmath4 classes is determined . a discusion on the size of these regions is also presented .
several years ago stillinger and helfand@xcite introduced a simple but nontrivial model of fluid demixing . their original model consists of a binary mixture of @xmath1 and @xmath2 particles . particles of the same type do not interact with one another , but @xmath1 and @xmath2 particles interact with a repulsive potential such that the mayer @xmath3-function is a gaussian . this choice for the @xmath4 potential , known as the gaussian molecule potential , greatly simplifies the calculation of virial coefficients and most work for this potential has been done using series methods@xcite . the main motivation for this work is to confirm ising universality for the critical exponents of continuum systems . in this paper we study the model and some of its generalizations using cluster monte carlo methods . where possible , we compare our results to the series analyses and to results for the ising - potts universality classes . although the gaussian molecule potential yields a more tractable virial expansion , it is easier to implement a cluster algorithm for the repulsive step potential . we also consider a generalization of the model to @xmath0 species ( components ) , such that particles of the same species do not interact but particles of different species interact with a repulsive potential . we expect that this generalization will be in the same universality class as the @xmath0-state potts model for @xmath0 not too large and another motivation for this study is to confirm this correspondence . for example , we know that the two - dimensional ( 2d ) potts model for @xmath5 has a first - order transition . does the @xmath0-component 2d model also have a first - order transition for @xmath5 ? in addition , we consider the effect of quenched disorder by randomly adding fixed scattering centers . there are general arguments@xcite that quenched disorder causes first - order transitions to become continuous . these arguments hold rigorously for 2d potts models@xcite , but have not been studied for continuum models . in previous work@xcite cluster monte carlo methods were applied to the widom - rowlinson model@xcite . the widom - rowlinson and models are closely related , the only difference is that the widom - rowlinson model has a hard - core interaction between different species . in this paper , the invaded cluster monte carlo method introduced in ref.@xcite is extended to soft - core repulsive potentials and is used to find the phase transition point for a given temperature without prior knowledge of the critical fugacity . the invaded cluster method has almost no critical slowing for the widom - rowlinson model , and we find that similar results hold for the models studied here . we consider @xmath0-component ( @xmath6 ) fluids in @xmath7 dimensions with @xmath8 . the components ( species ) have no self - interaction but particles of one species interact with particles of all other species via an isotropic repulsive potential , @xmath9 . we consider two choices for @xmath9 , [ eqn : allpots ] @xmath10 @xmath11 the limit @xmath12 for the step potential corresponds to the widom - rowlinson model . for the gaussian molecule potential , the temperature @xmath13 plays no role because the boltzmann factor , @xmath14 is , by design , independent of @xmath13 . in general , each component of the @xmath0-component fluid may have a distinct fugacity ; however symmetry considerations dictate that a demixing transition occurs with all fugacities equal , and hence we set all the fugacities equal to a single value , @xmath15 . for sufficiently small @xmath15 , the @xmath0 species are mixed , while for large @xmath15 , there are @xmath0 distinct phases because different species repel one another , with each phase predominately composed of one species . if @xmath0 is not too large , there is expected to be a single demixing transition separating these regimes that is in the same universality class as the @xmath0-state potts model . for very large @xmath0 , the correspondence between potts models and widom - rowlinson models must break down . although potts models have a single ordering transition , widom - rowlinson models can be presumed to have an intermediate crystalline phase for @xmath16 and large @xmath0 . to understand this phase , consider the limits @xmath17 and @xmath18 , with the product @xmath19 order unity . then non - overlapping particles appear with an effective fugacity of @xmath20 . however when two particles overlap , the cost is an additional factor of @xmath21 because the overlapping particles must be of the same species . hence the limiting model is precisely the hard sphere gas which we presume has a crystalline phase in @xmath16@xcite . it is therefore reasonable to assume that such a phase occurs in the widom - rowlinson models for large @xmath0 and @xmath15 of order @xmath22 . needless to say , for fixed @xmath0 , when the fugacity is sufficiently large , the model will demix , and thus the crystalline phase is an intermediate phase . we also expect an intermediate crystalline phase for large @xmath0 soft - core models based on a mapping to a single component fluid with a repulsive soft - core potential . for example , the gaussian core model@xcite is known to crystallize in @xmath23 . although intermediate phases do not occur for the usual potts models , they are not a consequence of the continuum ; indeed such phases are known to occur on the lattice for the site - dilute ( annealed ) potts models@xcite as well as for the lattice version of the widom - rowlinson model@xcite . the algorithms used here are , broadly speaking , examples of cluster algorithms of the type first introduced by swendsen and wang@xcite . cluster algorithms have been found to be much more efficient than local algorithms such as the metropolis algorithm for simulating spin systems and lattice gases near critical points . cluster algorithms would be very useful for off - lattice systems , but no general cluster method has yet been developed ; indeed , the _ only _ off - lattice models for which highly efficient cluster methods are known are models of the stillinger - helfand and widom - rowlinson type . the distinguishing features of this class of models are that particles of the same species have no self - interaction and that there is a purely repulsive interaction between particles of different species . in this case , graphical representations and clusters algorithms are available@xcite and have been implemented for the widom - rowlinson model@xcite . cluster algorithms for spin systems work by identifying clusters of spins and then randomly flipping these clusters . cluster are defined by placing bonds between nearest neighbor aligned spins with a probability that depends on the temperature . for fluid systems , bonds are placed between particles of the same type with a probability that depends on the temperature and the interaction potential . instead of flipping spins , clusters of particles are removed from the system and then new particles are added via a nonuniform poisson process that depends on the fugacity and the potential due to the remaining particles . cluster algorithms are typically used with fixed values of the external parameters such as temperature or fugacity . however , when the location of the phase transition is not known , much computational effort in studying the transition is spent in locating the transition . to avoid this problem , invaded cluster methods can be used@xcite which automatically adjust a thermodynamic parameter ( for example , temperature or fugacity ) to its value at the phase transition . this adjustment is accomplished by using the fact ( proved for the @xmath24 case@xcite ) that the clusters just percolate at the transition . in invaded cluster algorithms , clusters are grown until a signature of percolation is observed . the value of the thermodynamic parameter at the transition is an output of the simulation obtained from the fraction of successful attempts to add particles or bonds to the system . the invaded cluster algorithm also may be used to distinguish first - order from continuous transitions as discussed in ref.@xcite for potts models . this method for distinguishing the order of the transition is discussed below and will be used in section [ sec : stepq ] . we first describe the cluster algorithm discussed in section 3.5 of ref.@xcite for models and then discuss how it can be modified to be an invaded cluster algorithm . we assume that we have a configuration consisting of particle positions and a set of bonds connecting some of the particles and describe how to obtain the next configuration : 1 . identify all clusters of particles defined by the bonds . a particle with no bonds is considered to be a singleton cluster . for each cluster , independently and with probability @xmath21 , label it a _ black _ cluster and with probability @xmath25 label it _ white_. 2 . remove all particles in black clusters . the remaining white particles are at a set of positions @xmath26 . 3 . replenish the black particles via a poisson process with local intensity @xmath27 given by + @xmath28 + @xmath29 + where @xmath15 is the fugacity and @xmath9 the potential . 4 . for each pair of black particles , place a new _ bond _ between them with probability @xmath30 given by @xmath31 where @xmath32 is the separation between the particles . note that @xmath30 is minus the mayer @xmath3-function for the potential . eliminate the white and black labels for the clusters . this procedure comprises one monte carlo step . given a configuration of particle positions and bonds without species labels , it is possible to obtain a full multicomponent configuration where each particle has a species label . this assignment is accomplished by identifying clusters and then randomly and independently assigning one of the @xmath0 species labels to each cluster . the species label of each particle is taken to be the species label of its cluster . this labeling of particles is only possible if @xmath0 is a positive integer . however , the algorithm makes sense for all @xmath6 , in analogy to the relation between potts models , which are defined for positive integer @xmath0 , and random cluster models which interpolate between them and are defined for all @xmath33 . it is instructive to consider the nature of the cluster configurations generated by the algorithm as a function of the fugacity for a fixed temperature . suppose that the fugacity and , hence , the density is very small . then @xmath30 is typically small because the particles are far apart , and most clusters are singletons . in step 2 , a fraction @xmath21 of the particles is removed . in step 3 , particles are replenished as a nearly ideal gas because the exponential factor in eq . ( [ eq : intensity ] ) is a small perturbation except in the vicinity of the remaining particles . the end result is a nearly ideal multi - component gas . in the limit of large fugacity and density , we expect a phase in which a single species is predominant with a small admixture of the other species . the bonds connecting particles of the dominant species are sufficiently dense that almost all members of this species are in a single large cluster . the minority species are almost all in widely scattered singleton clusters . when the majority species is white , as occurs in about @xmath21 of the monte carlo steps , the large cluster is removed and then replaced as a nearly ideal gas in a slightly perturbed background potential generated by the minority species . an important feature of this picture is that the clusters do not percolate at small fugacity and do percolate at large fugacity . at some intermediate value of the fugacity , there must be a percolation transition . as discussed in refs.@xcite , the percolation transition of the clusters coincides with the demixing transition of the fluid . the coincidence of the percolation transition and the demixing transition justifies an invaded cluster version of the above cluster algorithm . the invaded cluster algorithm is very similar to the fixed @xmath15 cluster algorithm described above except that steps 3 and 4 are modified as follows . instead of putting down new black particles as a poisson process at a fixed intensity , black particles are added to the system one at a time according to the potential @xmath34 ( see below ) . after each black particle is added , bonds between the new particle and all previously placed black particles are put down with probability @xmath35 . the black clusters defined by these bonds are monitored after each particle is added , and the process of adding particles is stopped when a _ stopping condition _ is satisfied . for simulating the phase transition , the stopping condition is that one cluster _ spans _ the system . for periodic boundary conditions , spanning is taken to mean that a cluster wraps around the system in at least one of the @xmath7 directions . the spanning condition insures that the algorithm simulates the phase transition@xcite . in practice , a particle is added to the system according to the potential @xmath34 by the following procedure . a particle is tentatively placed at a random position @xmath36 . a random number @xmath32 is chosen in the interval @xmath37 , and the particle placement is accepted if @xmath38 otherwise the particle is rejected and another attempt is made to place a particle . let @xmath39 , where @xmath40 is the total number of attempted particle placements in a monte carlo step , including both accepted and rejected placements , @xmath41 is the system volume , and the brackets @xmath42 indicate an average over the simulation . because the intensity @xmath27 , defined in eq . ( [ eq : intensity ] ) , and the boltzmann factor @xmath43 governing particle placements differ by a factor of the fugacity , we conclude that @xmath44 is an estimator of @xmath45 , the value of the fugacity at the transition . note that if the fluctuations @xmath46 in @xmath47 are small , then the invaded cluster algorithm is essentially identical to the fixed fugacity algorithm operating at @xmath48 . this identification justifies the use of the invaded cluster method . a more complete discussion of the invaded cluster method and the use of @xmath44 as an estimator of a critical parameter is given in ref.@xcite . whenever the invaded cluster method simulates a system at its critical point , scaling methods can be used to obtain critical exponents from the size dependence of divergent thermodynamic quantities such as the compressibility or the susceptibility . to study the latter , we consider the quantity @xmath49 where @xmath50 is the number of particles in the @xmath51th cluster . we now show that @xmath52 is related to the usual susceptibility . consider , for simplicity , the discretized version of the model on a lattice of linear dimension @xmath53 with spacing @xmath54 so that the total number of sites is @xmath55^d$ ] . the demixing order parameter at site @xmath36 is given by @xmath56 , where @xmath57 if there is a particle of type 1 at site @xmath36 and @xmath58 otherwise ; @xmath59 counts the presence of a particle of _ any _ type . the relevant susceptibility @xmath60 is defined by the second derivative of the pressure with respect to the ( ordering ) chemical potential : @xmath61 ( the reason that @xmath54 does not enter explicitly into eq . ( [ eq : chidef ] ) is that the derivatives are with respect to the log of the activity and it is the activity that is scaled by @xmath54 . ) for a given particle and bond configuration , averaging over assignments of species labels , it is clear that @xmath62 vanishes unless the sites @xmath36 and @xmath63 are both occupied and in the same cluster , in which case the result is @xmath64 . thus , for a fixed particle - bond configuration , we obtain the number of particles in the cluster at @xmath36 if we sum over @xmath63 . summing over @xmath36 yields the sum of the squares of the cluster sizes so that @xmath65 , and hence we conclude that @xmath52 is related to the usual susceptibility . finally , finite size scaling predicts that @xmath66 so that the scaling of @xmath52 with system size can be used to extract the magnetic exponent @xmath67 . cluster methods also may be used to distinguish first - order from continuous transitions . for this purpose , a fixed density stopping rule is used . black particles are added to the system until the density @xmath68 reaches a fixed value and then @xmath44 is measured . in this way the canonical ensemble is simulated rather than the grand canonical ensemble . this procedure is done for a range of densities near the transition . if the transition is continuous , the fugacity is a strictly increasing function of @xmath68 . however , if the transition is first - order , then the fugacity does not increase monotonically with increasing @xmath68 in the coexistence region . why does the nature of the @xmath44 versus @xmath68 curve signify whether a transition is continuous or first - order ? suppose that the demixing transition of a @xmath0-component system is first - order . at the transition , there is coexistence of @xmath69 phases ; @xmath0 demixed phases and one mixed phase . because the repulsive interaction is reduced for the demixed phases , these phases have a higher density than the mixed phase . thus , in the thermodynamic limit , there is a range of @xmath68 for which the fugacity is constant . let @xmath70 be the density of the mixed phase and @xmath71 the density of the demixed phase . because @xmath72 , where @xmath73 is the entropy density , we have that @xmath73 is a linear function of @xmath68 in the coexistence region . more specifically , @xmath74 is a linear combination of @xmath75 and @xmath76 , the entropy densities of the mixed and demixed phases . the linearity of @xmath74 applies in the thermodynamic limit . however , for a finite system , the entropy density is not linear in the coexistence region . consider a system with linear dimension @xmath53 and periodic boundary conditions at density @xmath68 . this system also can be viewed as an infinite system with periodic constraints on the particles . let @xmath77 be the entropy density of this periodically constrained system . now suppose the constraints are removed and the system comes to equilibrium . if @xmath78 , demixing will occur spontaneously so that @xmath79 with the equality holding only at the endpoints of the coexistence range . because @xmath80 , we must have that @xmath15 is non - monotone in the coexistence region . this approach for distinguishing the order of a transition is very similar to the microcanonical monte carlo method used in ref.@xcite . in section [ sec : gaussian ] we present results for the 2d and 3d gaussian molecule models . the two - component step potential model is discussed in section [ sec : step2 ] and the @xmath0-component step potential is discussed in section [ sec : stepq ] . we simulated the gaussian molecule model ( with the potential @xmath81 defined in eq . ( [ eqn : sh ] ) ) using the invaded cluster method and the spanning rule described in section [ sec : methods ] for a range of linear dimensions @xmath53 up to 140 in @xmath82 and 40 in @xmath23 . we choose units such that distances are measured in units of @xmath83 . we collected statistics for the number of particles in the spanning cluster @xmath84 , the critical density @xmath68 , the susceptibility @xmath52 , the estimator of the critical fugacity @xmath44 and its standard deviation @xmath46 , and the normalized autocorrelation function for the spanning cluster size , @xmath85 . for each value of @xmath53 we averaged over @xmath86 monte carlo steps . the estimator of the critical density is the average number of particles ( of any species ) per unit area ( volume ) when the spanning condition is fulfilled . although @xmath87 does not go to zero at finite @xmath32 , it becomes very small for larger @xmath32 and to speed the calculation , we set @xmath88 for @xmath89 . tables [ table1 ] and [ table2 ] show the @xmath53 dependence of @xmath84 , @xmath68 , @xmath52 , @xmath44 , @xmath46 , and @xmath90 for the 2d and 3d models , respectively . the integrated autocorrelation time @xmath90 is defined by @xmath91 this time is approximately the number of monte carlo steps between statistically independent configurations and enters into the error estimate for @xmath84 . in practice , @xmath92 becomes indistinguishable from the noise for @xmath93 monte carlo steps , and it is necessary to cut off the upper limit of the sum defining @xmath90 when the magnitude of @xmath85 becomes comparable to its error . note that the fluctuations @xmath46 in @xmath44 decrease with increasing @xmath53 and that @xmath90 is small and hardly increases with @xmath53 . these results demonstrate the validity and efficiency of the invaded cluster algorithm . the decrease in @xmath46 shows that as @xmath53 increases , the invaded cluster becomes essentially equivalent to a fixed parameter cluster algorithm for which detailed balance can be proven . the error estimates for all quantities in tables [ table1 ] and [ table2 ] except @xmath90 were obtained by computing the standard deviation of the quantity of interest and dividing by the square root of the number of measurements . this error estimate does not take into account correlations between successive monte carlo steps . to account for correlations , the error estimates in the tables for an observable @xmath94 must be multiplied by @xmath95 , where @xmath96 is the integrated autocorrelation time for @xmath94 . the statistical errors for quantities derived from fits such as @xmath97 and @xmath98 include the factor @xmath99 except that @xmath96 is replaced by @xmath90 . figure [ fig : rho2d ] shows the results for @xmath100 versus @xmath101 for the 2d model . the value of @xmath68 in the limit of @xmath102 is extrapolated from the finite size data by doing a linear least squares fit omitting the values for @xmath103 and @xmath104 yielding the result , @xmath105 . a similar extrapolation for the critical fugacity yields @xmath106 . similarly , extrapolating the result for the 3d model using the data for all available @xmath53 yields @xmath107 and @xmath108 . our error values for these critical parameters are one standard deviation from the linear least squares fit of the fugacity or density versus @xmath101 ; no effort has been made to estimate systematic errors . all of the fits have acceptable goodness - of - fit probability values @xmath109 . our 3d value for the critical density is consistent with the series result of lai and fisher@xcite , @xmath110 ( eq . ( 36 ) of ref.@xcite ) but our critical fugacity is somewhat larger than their value , @xmath111 ( eq . ( 44 ) of ref.@xcite ) . note that lai and fisher report results using a different convention so that their values of @xmath97 and @xmath112 must be divided by @xmath113 to compare with our values . the exponent ratio @xmath98 can be obtained from the scaling of the susceptibility @xmath52 with @xmath53 according to eq . ( [ eq : chifss ] ) . figure [ fig : chi2 ] shows a log - log plot of @xmath52 versus @xmath53 for the 2d gaussian molecule model . a least squares fit of all the data to a simple power law does not yield an acceptable goodness of fit value @xmath109 . if the smallest value of @xmath53 is omitted , we obtain @xmath114 with @xmath115 , @xmath116 , and @xmath117 ( degrees of freedom ) . the @xmath109 value indicates a reasonable fit to a simple power law , but the fitted value of @xmath98 is @xmath118 from the 2d ising value of @xmath119 . a reasonable explanation of this result is that the 2d gaussian molecule is , indeed , in the 2d ising universality class , but that there are relatively slowly varying corrections to scaling . a least squares fit to all the data for the susceptibility @xmath52 for the 3d gaussian molecule model yields @xmath120 with @xmath121 , @xmath122 , and @xmath123 . the @xmath109 value near unity suggests that the data is well fit to a pure power law . recent high precision monte carlo studies of the 3d ising model@xcite yield @xmath124 which is consistent with our results . our results add weight to the hypothesis that the model is in the ising universality class for both 2d and 3d . the relatively high precision results for @xmath98 from the gaussian molecule model suggests that models of the type may be useful for high precision studies of the 3d ising universality class . the isotropy of the interaction and absence of an underlying lattice might make for smaller corrections to scaling in models compared to lattice spin models . table [ table3 ] summarizes our results for the 2d model with the step potential given by eq . ( [ eqn : step ] ) and temperature @xmath125 ( measured in units of @xmath126 ) . for each value of @xmath53 we averaged over @xmath127 monte carlo steps . the results are qualitatively similar to the gaussian molecule model . table [ table4 ] shows the temperature dependence of the measured quantities for @xmath53 fixed at @xmath103 . the values of @xmath68 and @xmath44 at low temperatures should reduce to the widom - rowlinson model . in ref.@xcite we measured the critical parameters of the widom - rowlinson model using the invaded cluster method . for @xmath128 ( the smallest size measured ) we obtained @xmath129 and @xmath130 , values that are close to the values of @xmath68 and @xmath44 for the two lowest temperatures in table [ table4 ] this agreement confirms that the step potential is continuously connected to the hard core potential . if the @xmath103 data point is omitted , we obtain from a least squares fit to the data for the susceptibility @xmath52 , @xmath131 with @xmath132 , @xmath133 , and @xmath134 . [ sec : stepq ] the critical properties of the @xmath0-component model are expected to be closely related to the @xmath0-state potts model . one of the features of the @xmath0-state potts model is that the transition is continuous for small @xmath0 and is first - order for @xmath135 , where @xmath136 and @xmath137 . we have used the method described in section [ sec : methods ] to determine the order of the transition as a function of @xmath0 for the @xmath0-component step potential model . figure [ fig : fig4 ] shows the fugacity @xmath44 as a function of @xmath68 for @xmath138 for @xmath128 and @xmath125 . note that for @xmath139 the curve is clearly monotonically increasing , which implies a continuous transition . for @xmath140 the curves are clearly non - monotonic , which implies a first - order transition . for @xmath141 the curve is essentially flat within the error bars ( whose size is approximately that of the symbols ) . although the effective value of @xmath142 is expected to vary with @xmath53 , these results are consistent with the hypothesis that @xmath136 for the 2d step potential model . figure [ fig : fig5 ] shows @xmath44 as a function of @xmath68 for the 3d step potential model for @xmath103 and @xmath125 . the @xmath24 curve is clearly monotonically increasing while the @xmath139 is clearly not , implying that @xmath137 as for the 3d potts model . finally , we have studied the effect of quenched impurities on the nature of the transition for the @xmath139 step potential model . the impurities consist of randomly placed scatterers that interact with all the fluid particles via the same repulsive step potential that exists between different components . figure [ fig : fig6 ] shows a plot of @xmath44 versus @xmath68 for the @xmath139 step potential model in 3d for four impurity densities ranging from 0.025 to 0.0625 . for each of the 10 impurity configurations considered for a given density , data from @xmath143 monte carlo steps are collected . for the two lowest impurity concentrations , the @xmath44 versus @xmath68 curve is non - monotonic as is the case for the pure system , while for the two highest impurity concentrations , the curve is monotonic indicating a crossover to a continuous transition . this behavior is in accord with general arguments@xcite that the presence of quenched impurities should cause a first - order transition to become continuous . it is not clear from our data whether there is a critical value of the disorder below which the transition remains continuous or whether the crossover at finite disorder strength is a finite size effect and that any strength of disorder is sufficient to make the transition continuous in the thermodynamic limit . we have studied the model and several generalizations using the invaded cluster algorithm . our results for @xmath0-component models with @xmath144 are consistent with the hypothesis that these models are in the same universality class as the corresponding potts models . in addition , we have shown that the addition of quenched disorder causes the demixing transition to change from first - order to continous for those values of @xmath0 for which the pure system transition is first - order . for the case @xmath24 and @xmath138 , our results for the magnetic exponent are outside the statistical error bars of the exact ising value . however , we believe that this difference is most likely the result of slowly varying corrections to scaling . it would be useful to consider larger system to confirm ising universality . it would also be interesting to consider larger values of @xmath0 to explore the possibility of an intermediate crystalline phase in models . this work was supported by nsf grants phy-9801878 ( rs ) , dmr-9633385 ( hg ) , dmr-9978233 ( jm ) , dms-9971016 ( lc ) , and nsa grant mda904 - 98 - 1 - 0518 ( lc ) . we thank gregory johnson for useful discussions . .the @xmath53 dependence of the number of particles in the spanning cluster @xmath84 , the critical density @xmath68 , the susceptibility @xmath52 , the estimator of the critical fugacity @xmath44 , its standard deviation @xmath46 , and decorrelation time @xmath90 for the 2d gaussian molecule potential . the averages are over @xmath146 spanning clusters . the error estimates were obtained by computing the standard deviation of the quantity of interest and dividing by the square root of the number of measurements . the error estimates for @xmath90 are obtained from the variation of @xmath147 with the upper limit in the summation of eq . ( [ eq : tau ] ) and hence represent an estimate of the systematic error . the autocorrelation function @xmath92 is distinguishable from the noise for @xmath148 monte carlo steps . [ cols=">,^,^,^,^,^,^ " , ]	phase transitions of fluid mixtures of the type introduced by stillinger and helfand are studied using a continuum version of the invaded cluster algorithm . particles of the same species do not interact , but particles of different types interact with each other via a repulsive potential . examples of interactions include the gaussian molecule potential and a repulsive step potential . accurate values of the critical density , fugacity and magnetic exponent are found in two and three dimensions for the two - species model . the effect of varying the number of species and of introducing quenched impurities is also investigated . in all the cases studied , mixtures of @xmath0-species are found to have properties similar to @xmath0-state potts models .
the study of _ diffusion _ in nearly - integrable hamiltonian dynamical systems of the form @xmath5 where @xmath6 are n - dimensional action - angle variables and @xmath0 is a small parameter , constitutes a central problem in hamiltonian dynamical systems theory , in view , in particular , of its multiple applications in physics and astronomy ( see @xcite@xcite for an introduction , the basic review paper @xcite , or @xcite@xcite @xcite for recent advanced reviews emphasizing various aspects of this subject ) . it is a well established result that , if @xmath7 , and @xmath8 satisfies appropriate convexity and analyticity conditions ( see section 2 below ) , two distinct regimes characterize the laws of diffusion as a function of @xmath0 : for @xmath9 , where @xmath10 is a threshold value , the onset of the so - called ` nekhoroshev regime ' takes place @xcite@xcite@xcite @xcite@xcite@xcite . in this case , the nekhoroshev theorem provides an @xmath11 $ ] upper bound for the speed of diffusion . the exponent @xmath12 depends on the number of degrees of freedom @xmath13 , while its precise value in local domains of the action space depends also on the multiplicity of the resonance conditions holding in such domains ( see e.g. @xcite@xcite@xcite ) . furthermore , the mechanism of diffusion caused by _ transition chains _ , as demonstrated in one special example by arnold @xcite ( see also @xcite ) , is conjectured to hold in more general systems of the form ( [ hamgen ] ) ( e.g. @xcite ; note , however , that no formal proof of this fact has been given to date ) . on the other hand , for @xmath14 , the diffusion is driven mainly by the mechanism of _ resonance overlap _ @xcite@xcite @xcite . in this case , one expects a power - law dependence of the speed of diffusion on @xmath0 ( see e.g. @xcite ; a power law is also found in the case of the so - called ` fast arnold diffusion ' @xcite ) . the diffusion in weakly chaotic systems has been a subject also of extensive numerical studies over the last three decades ( some indicative references are @xcite@xcite@xcite@xcite @xcite@xcite@xcite@xcite ) . a detailed study , however , of the very slow diffusion characterizing the ` nekhoroshev regime ' has become possible only in recent years . in this respect , we note in particular the series of instructive works @xcite @xcite@xcite@xcite@xcite , where , using the so - called fast lyapunov indicator ( fli ; see @xcite ) , a method was found to depict the resonant structure of the action space in models of three degrees of freedom , or 4d and 6d symplectic mappings being in the nekhoroshev regime @xcite . in @xcite , the mean - square spread in action space @xmath15 was measured as a function of the time @xmath16 for orbits along the chaotic border of a _ simply - resonant _ domain ( see section 2 for a precise definition ) . it was found that i ) the local character of diffusion is normal , i.e. @xmath17 , and ii ) the diffusion coefficient @xmath18 decreases with @xmath0 faster than a power law . the exponential fit @xmath19 was given in a subsequent study @xcite . the estimate obtained in @xcite , through interpolation over five orders of magnitude of the perturbation parameter , yields with with certainty the first digit of the exponent @xmath20 , but the errors in the interpolation make uncertain the second digit in both the above estimates . in @xcite , @xmath1 was measured as a function of the separatrix splitting @xmath21 of the asymptotic manifolds of simply unstable two - dimensional tori lying at the borders of simple resonances ( see also @xcite@xcite ) . the measurement of @xmath21 itself was based on employing the fli . it was found that @xmath22 , with @xmath23 and @xmath24 in two resonances of increasing order respectively . finally , the laws of diffusion in systems violating one or more necessary conditions of the nekhoroshev theorem were investigated in @xcite@xcite , leading to a number of interesting results regarding the dynamical consequences of such violations . the motivation for the present study stems primarily from the results reported in refs @xcite@xcite @xcite @xcite , and it can be described as follows . the results obtained so far are very satisfactory from the numerical point of view . they require , however , computations involving large ensembles of orbits and integration times of the order of billions , or even trillions of periods . on the other hand , we can remark that , in principle , the analytical methods involved in the main theories of chaotic diffusion lend themselves also conveniently to getting quantitative predictions regarding the value of the diffusion coefficient , or the scaling laws of diffusion , in general , in the weakly chaotic regime . for such a goal , however , to be accomplished , it is required that one should be able to carry on expansions of certain quantities up to a very high order in the small parameter @xmath0 ( usually with the aid of a computer ) . this fact is explicit in nekhoroshev theory , where one needs to reach an expansion order high enough for the asymptotic behavior of the perturbation series to show up . this has been realized in studies seeking to determine the _ range _ ( in the small parameter value ) and/or the _ conditions of applicability _ of nekhoroshev theory , or , finally , the _ domain of practical stability _ for motions in simple physical systems or models inspired mainly from solar system dynamics ( see e.g. @xcite@xcite @xcite@xcite@xcite@xcite @xcite@xcite@xcite ) . these studies notwithstanding , the question of central interest in the present paper , namely how to obtain relevant quantitative estimates of the _ local value of the diffusion coefficient d _ in resonant domains ( of various multiplicities ) of the action space via _ high order expansions _ of perturbation theory , remains , to our knowledge , largely unexplored . regarding now this last question , it should be noted that the formal analytical apparatus of nekhoroshev theory , entailing the construction of a _ normal form _ in local domains covering the action space of systems of the form ( [ hamgen ] ) , aims to transform the original hamiltonian into one in new canonical variables resuming the form @xmath25 , where @xmath26 , the normal form , corresponds to a simple dynamics , while @xmath27 , the remainder , induces a perturbation to this dynamics . the so - called ` geometric part ' of nekhoroshev the theorem ensures that , despite allowing in general for chaotic motions , the flow under a multiply - resonant normal form _ alone _ would imply perpetual confinement of all chaotic orbits in balls of radius @xmath28 in the action space . nevertheless , this picture is altered due to the effects of the remainder which eventually causes the orbits to diffuse away of their initial @xmath28 domain . now , via a sequence of hamiltonian normalization steps we find that there is an optimal order at which the size of the remainder becomes exponentially small in a power of @xmath29 . this , in turn , implies an exponentially small _ semi - analytic upper bound _ of the value of the diffusion coefficient @xmath1 . unfortunately , such a bound turns usually to be very unrealistic , as it overestimates by a large factor the true value of @xmath1 ( or , equivalently , it underestimates the time of practical stability ) . we are thus led to conclude that , whereas the remainder @xmath27 constitutes a quantity of primary interest in quantitative applications of nekhoroshev theory , the precise relation between @xmath27 and @xmath1 is apparently very different from what upper bound estimates would suggest . instead , a detailed analysis of the _ effects of the remainder on dynamics _ appears to be necessary in order to formulate a more precise theory of the relation between @xmath27 and @xmath1 . in the sequel , we present such an analysis in systems of three degrees of freedom . in this analysis , we still have to rely on an assumption for which numerical indications are available , namely that the local character of diffusion in sufficiently small domains of the action space is ` normal ' , that is , the mean square spread of the actions of the chaotic orbits grows linearly with time ( there are indications that _ global diffusion _ , which concerns ensembles or orbits diffusing in a substantial part of the arnold web over much longer timescales , could also be described as ` normal ' ( see @xcite ) ; however , the issue of the laws of global diffusion can only be hoped to tackle after the laws of local diffusion have been adequately understood ) . in the rest of our analysis , we proceed by expressing all quantities of interest in terms of the remainder function , which , in turn , is calculated in concrete examples by a well - defined algebraic procedure . finally , we estimate via this analysis how @xmath1 depends on the size @xmath2 of the remainder at the optimal normalization order . it should be noted that the idea that the stability properties of the orbits in nearly - integrable systems depend on the size of the optimal remainder is not new , but it is one permeating nearly all forms of canonical perturbation theory . the novel feature here , instead , is to use @xmath2 not as an upper bound for @xmath1 , but as a way to estimate @xmath1 via examining the relation between the two quantities as determined by independent numerical experiments . one main prediction is that this relation is altered according to the _ multiplicity of resonance conditions _ holding in the action domain of interest . more concretely , we predict that the diffusion coefficient @xmath1 scales with @xmath2 as a power - law @xmath30 in simply resonant domains , for some constant @xmath31 . a combination of theoretical arguments found in @xcite@xcite , together with quantitative estimates on the relation between the size of the so - called separatrix splitting ( see subsection 2.3.2 ) and the normal form remainder given in @xcite , suggest @xmath32 , i.e. @xmath33 in simply resonant domains . this agrees with the numerical results obtained in a previous study @xcite . in @xcite , a computer - algebraic program was written in order to calculate the optimal normal form as well as the remainder function @xmath34 at the optimal mormalization order in a case of simple resonance , employing the same hamiltonian model as in @xcite . this operation involved computing about @xmath35 fourier coefficients , at a truncation order in fourier space as high as @xmath36 . comparing the computed size of @xmath2 versus available numerical data on @xmath1 from @xcite , the scaling @xmath37 was found by numerical fitting . in the present paper , after presenting some theoretical results , we make a similar numerical calculation as in @xcite but in the case of a double resonance . in order to reach the optimal normalization , we had to extend all normal form calculations up to the fourier order @xmath38 ( @xmath39 coefficients ) . we thus determined the size of the optimal remainder @xmath2 for many different values of the small parameter @xmath0 . in the same time , we computed the diffusion coefficient @xmath1 for the same values of @xmath0 by a purely numerical procedure involving runs of ensembles of chaotic orbits ( see section 3 ) . finally , we made two independent numerical comparisons of the relation between @xmath1 and @xmath2 . the latter yield the power laws @xmath40 is close to 2 in doubly resonant domains , albeit with a small noticeable difference even in this case , which probably requires a more precise theory to interpret . besides the above computation , our analysis using high order normal forms resulted in a relevant result regarding the possibility to visualize how the phenomenon of arnold diffusion proceeds locally , within a doubly - resonant domain , by materializing the computation of a convenient set of variables helping to this purpose , that were proposed in the work @xcite . we note that numerical evidence for arnold diffusion of orbits entering from simple to double resonances was presented in @xcite . here , we provide a detailed topological description of this phenomenon . the whole computation consists of : i ) computing a set of resonant canonical action - angle variables via a sequence of lie canonical transformations , ii ) taking a 2d poincar surface of section of the _ doubly - resonant normal form dynamics _ ( which represents a system of two degrees of freedom ) , and ( more importantly ) iii ) using the _ energy @xmath41 of the normal form _ as the third variable , showing the effect of arnold diffusion . according to theory , the value of @xmath41 changes exponentially slowly in time due to the effect of the remainder . in the sequel we refer to this phenomenon as ` drift ' , although in reality it means that a number of quantities can be characterized as undergoing random walk during the whole diffusion process . besides setting the timescale of diffusion , the drift can be viewed also as the source of a dynamical phenomenon , namely the communication between chaotic domains that would be otherwise isolated under the doubly - resonant normal form hamiltonian flow . we show in a true example the excursion of a chaotic orbit within the doubly - resonant domain as it appears in the above proposed set of variables . we thus identify a sequence of chaotic transitions of such an orbit from one resonant domain to another . in fact , in each transition the orbit bypasses the barriers imposed by normal form dynamics via a ` third dimension ' , i.e. the slowly drifting value of @xmath41 . we finally argue that , besides their practical utility , such illustrations are also suggestive of the geometric structure underlying the asymptotic manifolds of lower - dimensional tori filling the phase space in the domain of a double resonance . these manifolds are important , because , following the spirit of arnold s original work @xcite , it has been widely conjectured that their heteroclinic intersections constitute a primary cause of arnold diffusion . of course , proving this fact represents a well known important open problem of dynamical systems theory . the structure of the paper is as follows : section 2 presents the theory , focusing on the normal form algorithm , multiply - resonant dynamics , effect of the remainder , and , finally , on the relation between @xmath1 and @xmath2 . we describe in some length all necessary theoretical steps in order to render the paper as self - contained as possible . section 3 then passes to the numerical results . we present i ) the results from the normal form computer - algebraic construction , ii ) the visualization of arnold diffusion using appropriate variables based on the normal form computation , iii ) the numerical calculation of the diffusion coefficient @xmath1 , and , finally iv ) the comparison of @xmath1 with @xmath2 . section 4 summarizes the main conclusions of the present study . most statements made in subsections 2.1 and 2.2 below , regarding the properties of the hamiltonian models considered as well as the algorithm by which we perform hamiltonian normalization , are applicable to systems of an arbitrary number of degrees of freedom . in order , however , to be consistent with the rest of the paper , we use everywhere a notation referring to systems of three degrees of freedom . on the other hand , the analysis of subsection 2.3 applies to the study of the diffusion in doubly or simply resonant domains . in systems of three degrees of freedom , the latter represent the only possible multiplicities of a resonance condition , while in systems of more than three degrees of freedom there are also cases of intermediate resonance multiplicities between one and the maximal . the latter s study , nevertheless , is well beyond our present computational capacity , and thus it is left as an open problem . we consider three degrees of freedom systems of the form ( [ hamgen ] ) , where @xmath8 satisfies the following analyticity and convexity conditions : \i ) _ analyticity : _ @xmath8 is assumed to be an analytic function in a complexified domain of its arguments . namely , we assume that there is an open domain @xmath42 and a positive number @xmath43 such that for all points @xmath44 and all complex quantities @xmath45 satisfying the inequalities @xmath46 , the function @xmath47 admits a convergent taylor expansion @xmath48 where @xmath49 , and @xmath50 are the entries of the hessian matrix of @xmath47 at @xmath51 . furthermore , we assume that there is a positive constant @xmath52 such that for all @xmath53 , @xmath54 admits an absolutely convergent fourier expansion @xmath55 in a domain where all three angles satisfy @xmath56 , @xmath57 . by the fourier theorem ( see e.g. @xcite ) , this condition implies that the coefficients @xmath58 decay exponentially with the @xmath59modulus @xmath60 , that is , there is a positive constant @xmath61 such that the bound @xmath62 . we finally assume that all coefficients @xmath63 admit taylor expansions with respect to @xmath51 @xmath64 ( where @xmath65 are the entries of the hessian matrix of @xmath58 at @xmath51 ) , which are convergent in the same union of domains as for @xmath47 . \ii ) _ convexity : _ for the hessian matrix @xmath66 , which is real symmetric , we assume a simple quasi - convexity condition , namely that for all @xmath67 either two of the ( real ) eigenvalues of @xmath66 have the same sign and one is equal to zero , or all three eigenvalues have the same sign . furthermore , we define two constants : @xmath68 where @xmath69 is a label of only non - zero eigenvalues @xmath70 of @xmath66 , i.e. @xmath71 or @xmath72 if there are two or three non - zero eigenvalues respectively . as will be discussed in detail in subsection 2.3 , the quasi - convexity condition is essential , since it introduces a confinement of the orbits for exponentially long times on a surface arising from the condition of preservation of the energy ( see @xcite ) . we now give some definitions allowing to characterize resonant dynamics . a _ resonant manifold _ @xmath73 associated with a non - zero wavevector @xmath74 with co - prime integer components @xmath75 is the two - dimensional locus defined by @xmath76 where @xmath77 . let @xmath67 be such that all three frequencies @xmath78 , @xmath79 are different from zero . we now distinguish the following three cases : + i ) _ non - resonance : _ no resonant manifold @xmath73 contains @xmath51 . \ii ) _ simple resonance : _ one resonant manifold @xmath73 contains @xmath51 . \iii ) _ double resonance : _ more than one resonant manifolds contain @xmath51 . in the latter case , it is possible to choose two linearly independent vectors @xmath80 such that all resonant manifolds @xmath81 containing @xmath51 are labeled by vectors @xmath74 which are linear combinations of the chosen vectors @xmath80 with rational coefficients . the intersection of these manifolds forms a one - dimensional _ resonant junction_. a doubly - resonant point @xmath51 always corresponds to the intersection of a resonant junction with a constant energy surface @xmath82 . + in the above definitions , resonant manifolds @xmath73 of all possible wavevectors @xmath74 have been considered . it is well known , however , that in normal form theory a natural truncation limit @xmath83 arises in fourier space ( see below ) . accounting for this possibility , we call a point @xmath67 i ) non - resonant , ii ) simply resonant , or iii ) doubly resonant _ with respect to a k truncation _ , if the number of resonant manifolds @xmath73 with @xmath83 passing through @xmath51 are i ) zero , ii ) one and iii ) more than one respectively . finally , it will be convenient to introduce a definition concerning _ open domains _ in @xmath84 . let @xmath85 be a ball of radius @xmath86 around one point @xmath51 in @xmath84 . if @xmath47 satisfies convexity conditions as assumed above , for @xmath86 small whatsoever the domain @xmath85 is crossed by a dense set of resonant manifolds @xmath73 . however , for any fixed value of the positive integer @xmath87 , only a finite subset of the manifolds @xmath73 satisfy @xmath83 . the domain @xmath85 is then called : i ) non - resonant , ii ) simply - resonant , and iii ) doubly - resonant with respect to the @xmath87truncation if @xmath51 is , respectively , non - resonant , simply - resonant or doubly - resonant , and no other resonant manifolds @xmath73 with @xmath83 cross @xmath85 except for the ones passing through @xmath51 . all our estimates on the speed of diffusion are based on an appropriate normal form construction . in this , we adopt the method exposed in detail in @xcite , which lends itself conveniently to i ) developing a computer - algebraic program , and ii ) deriving analytical estimates on the size of various quantities appearing in the course of hamiltonian normalization . the main elements of this method are : + _ expansion centers . _ the action space can be covered by domains @xmath85 , centered around points @xmath51 which serve as expansion centers of both the original hamiltonian and the normal form . we choose the points @xmath51 to belong to the set of all doubly - resonant points of @xmath84 , denoted by @xmath88 , and by setting @xmath86 as of order @xmath28 . the covering is possible because @xmath88 is dense in @xmath84 . a normal form construction as done below is valid within one domain @xmath85 ( this is essentially the same starting point as in lochak s @xcite analytic construction leading to a proof of the nekhoroshev theorem ) . a crucial remark is that the characterization of dynamics within @xmath85 as non resonant , simply resonant , or doubly resonant depends on @xmath0 . this is because , as shown below , the optimal normal form truncation order @xmath89 in fourier space depends on the value of @xmath0 . furthermore , for a given value of @xmath87 , the set @xmath88 can be decomposed in three disjoint sets @xmath90 , containing all non - resonant , simply resonant and doubly resonant points respectively with respect to the @xmath87truncation . thus , the characterization of resonant dynamics within @xmath85 depends on whether , according to the value of @xmath87 , @xmath51 belongs to @xmath91 , @xmath92 , or @xmath93 . + _ resonant module : _ let @xmath51 be a point of @xmath88 and @xmath94 , @xmath95 two linearly independent vectors such that @xmath96 for @xmath97 . more than one choices of @xmath98 and @xmath99 are possible . in the sequel we choose @xmath98 and @xmath99 so that @xmath100 is minimal . the vector @xmath101 defined by @xmath102 is parallel to the vector @xmath103 since @xmath104 for all @xmath74 satisfying @xmath105 . if @xmath106 , @xmath107 , @xmath108 are not co - prime integers , we re - define @xmath109 by dividing the @xmath110 by their greatest common divisor . the set @xmath111 is hereafter called the resonant module associated with the point @xmath112 . the resonant module includes wavevectors @xmath74 whose respective trigonometric terms @xmath113 are to be retained in the normal form . + _ action re - scaling : _ from now on we focus on the construction of the normal form in one specific domain @xmath85 . it has been mentioned already that it is convenient to choose @xmath86 as a quantity scaling proportionally to @xmath114 . the simplest way to accommodate such a choice is by introducing the following re - scaling of all action variables within @xmath85 : @xmath115 this re - scaling greatly simplifies the normal form algorithm , because it formally removes all terms besides linear in the actions from the kernel of the so - called homological equation ( see below , or @xcite for details ) by which the normalizing generating functions are determined . eq.([resc ] ) does not define a canonical transformation . however , the correct equations of motion in the variables @xmath116 are produced by the hamiltonian function @xmath117 , i.e. ( neglecting a constant ) @xmath118 where the first line in the above equation comes from the integrable part @xmath47 of the original hamiltonian ( eq.([h0exp ] ) ) , while the second line comes from the perturbation @xmath54 ( eq.([h1four ] ) ) given the series expansion of the fourier coefficients as in eq.([hkexp ] ) . + _ book - keeping : _ we now split the hamiltonian ( [ hamexp2 ] ) in parts of different order of smallness , which are to be normalized step by step . the function ( [ hamexp2 ] ) contains terms of various orders in the small parameter @xmath114 . however , the presence of a second ` small parameter ' @xmath119 is implied in ( [ hamexp2 ] ) by the exponential decay of all fourier coefficients @xmath120 , @xmath65 , etc . , due to eq.([anal ] ) ( see @xcite pp.90 - 91 for a thorough exposition of the role of this small parameter in nekhoroshev theory ) . we take both parameters into account by introducing an integer @xmath121 such that @xmath122 , i.e. by setting : @xmath123~~.\ ] ] using @xmath121 , the hamiltonian ( [ hamexp2 ] ) can be split in groups of practically the same order of smallness . this is realized by artificially introducing a ` book - keeping ' coefficient @xmath124 in front of each term in ( [ hamexp2 ] ) , whose numerical value is set equal to unity at the end of the calculation . furthermore , for a term of the form @xmath125 we set @xmath126+\mu$ ] . regarding the above ` book - keeping ' process it is worth noting the following : i ) this way of splitting the hamiltonian in different orders of smallness results in a finite number of terms appearing in every power of @xmath127 . ii ) this technique is suggested already by poincar @xcite and arnold @xcite . in fact , the dependence of @xmath121 on @xmath0 is weak , since it is logarithmic , so that an alternative choice to the ` ansatz ' ( [ kprime ] ) is to set @xmath128 . in fact , according to giorgilli @xcite this is an optimal choice . iii ) since , at every normalization order , we have a reduction of the analyticity domain , one could consider re - defining @xmath121 at every normalization step . however , this is hardly tractable from an algorithmic point of view . instead , keeping @xmath121 constant at all normalization orders should be viewed as a rule indicating the sequence by which the various terms in the hamiltonian are normalized , i.e. , the terms or order @xmath129 are normalized in the r - th step . albeit not necessarily optimal regarding the grouping of the terms according to their size , this rule proves simple to implement and sufficient in practice . returning to the form of the hamiltonian after introducing the book - keeping factor @xmath127 , the hamiltonian reads : @xmath130}\epsilon^{1/2}h_{k}\\ & + & \lambda^{2+[\|k\|/k']}\epsilon\nabla_{i_}h_k\cdot j + \lambda^{3+[\|k\|/k']}{\epsilon^{3/2}\over 2 } \sum_{i=1}^3\sum_{j=1}^3h_{k , ij}j_ij_j + \ldots\bigg ) \exp(ik\cdot\phi)\nonumber~~.\end{aligned}\ ] ] + setting @xmath131 , the hamiltonian ( [ hamexp3 ] ) resumes the form @xmath132 where i ) the superscript @xmath133 denotes zeroth - step of the normalization procedure (= original hamiltonian ) , ii ) the exponent of @xmath127 in different terms keeps track of their true order of smallness , and iii ) the functions @xmath134 are of the form @xmath135 where @xmath136 are polynomials containing terms of degree @xmath137 or @xmath138 in the action variables @xmath139 . precisely , we have : @xmath140 if @xmath141 , or @xmath142 @xmath143 if @xmath144 . + _ hamiltonian normalization : _ we use the algorithm of composition of lie series in order to perform the hamiltonian normalization . let us recall that the purpose of the normalization is to introduce a sequence of canonical transformations @xmath145 @xmath146 @xmath147 so that the hamiltonian expressed as a function of the new variables allows one to more easily identify the main features of dynamics . after @xmath148 normalization steps , the old variables @xmath149 are expressed in terms of the new variables @xmath150 , and the hamiltonian @xmath151 takes the form @xmath152 the terms @xmath153 and @xmath154 are called the _ normal form _ and the _ remainder _ respectively . the normal form is a finite expression which contains terms up to order @xmath148 in the book - keeping constant @xmath127 , while the remainder is a series containing terms of order @xmath155 and beyond . the mathematical structure of the normal form term @xmath156 is such as to imply an easily identifiable dynamics in the variables @xmath157 ( e.g. an oscillator or pendulum dynamics ) . on the other hand , the remainder is a _ convergent _ series in a restriction of the domain of analyticity of the original hamiltonian , which represents a perturbation with respect to the hamiltonian flow of @xmath156 . an optimal normalization order @xmath158 exists ( see below ) where the process must be stopped . the hamiltonian normalization is implemented step - by - step by the recursive equation : @xmath159 where @xmath160 is the r - th step lie generating function and @xmath161 is the poisson bracket operator . both @xmath162 and @xmath160 are functions of the variables @xmath163 . the generating function is defined by the solution of the homological equation @xmath164 where @xmath165 denotes all terms of @xmath166 which i ) have a book - keeping coefficient @xmath129 in front , and ii ) belong to the range of the operator @xmath167 . given the definition of the resonant module @xmath168 in eq.([resmod ] ) , one has the relation @xmath169 where @xmath170 are all the terms of @xmath166 having a factor @xmath129 , and @xmath171 are the _ normal form _ terms of @xmath170 , that is all the trigonometric terms whose wavevectors @xmath74 belong to @xmath168 . it follows immediately that @xmath162 has the form @xmath172 where all terms in the functions @xmath173 have a factor @xmath174 , while @xmath175 is a series in powers of @xmath127 starting with terms of order @xmath155 . + _ optimal truncation : _ in the analytical part of the nekhoroshev theory it is demonstrated that the whole normalization process has an asymptotic character . namely , i ) the domain of convergence of the remainder series @xmath175 shrinks as the normalization order @xmath148 increases , and ii ) the size @xmath176 of @xmath175 , where @xmath177 is a properly defined norm in the space of trigonometric polynomials ( see below ) , initially decreases , as @xmath148 increases , up to an optimal order @xmath158 beyond which @xmath176 increases with @xmath148 . in the nekhoroshev regime , one has @xmath178 . thus , stopping at @xmath158 best unravels the dynamics , which is given essentially by the hamiltonian flow of @xmath179 slightly perturbed by @xmath180 . the long term consequences of this perturbation , which determine the speed of diffusion , will be analyzed in subsection 2.3 . the normal form @xmath181 contains trigonometric terms @xmath113 of order not greater than @xmath182 . let @xmath158 be the optimal normalization order . it is well known that the dependence of @xmath158 on @xmath0 is given by an inverse power - law , namely @xmath183 the exponents @xmath184 , @xmath185 and @xmath186 , referring to the non - resonant , simply resonant , and doubly resonant normal form constructions respectively , are found in @xcite . we emphasize that , while , due to the introduction of the book - keeping process , the algorithm of hamiltonian normalization analyzed above is not technically identical with the usual normalization procedure used in the proof of the nekhoroshev theorem ( e.g. as in @xcite ) , in practice we recover the estimate ( [ ropt ] ) , and the resulting exponents , both in the simply resonant case ( see @xcite ) and in the doubly resonant case , as confirmed by numerical experiments in section 3 below . in particular , we find that since the leading terms in the remainder are @xmath187 , the size of the remainder is of order @xmath188 , implying ( viz.eq.([kprime ] ) ) : latexmath:[\[\label{remropt } i.e. the remainder is exponentially small in @xmath29 in accordance with the nekhoroshev theorem . the fourier order @xmath190 is hereafter called the optimal k truncation order . all the normal form terms of @xmath191 have fourier orders satisfying @xmath192 . + we are now in a position to discuss the essence of all the previous definitions . the key point is to observe that , depending on the value of @xmath0 , _ the same _ expansion point @xmath112 of the normal form construction turns to be either non - resonant , or simply or doubly resonant with respect to the optimal k truncation . in particular , let @xmath98 and @xmath99 be two linearly independent vectors of @xmath168 such that for all @xmath193 one has @xmath194 . we then distinguish the following three regimes : i ) @xmath195 . then , the point @xmath51 is doubly - resonant with respect to the optimal k truncation . this is the case we mainly focus on in the sequel . the main theoretical results are given in subsection 2.3.1 , while the main numerical results are given in section 3 . ii ) @xmath196 . then , @xmath51 is simply - resonant with respect to the optimal k truncation . one such example was dealt with in the numerical study @xcite . further theoretical analysis of this case is made in subsection 2.3.2 . iii ) @xmath197 . then , @xmath51 is non - resonant with respect to the optimal k truncation . since @xmath198 decreases as @xmath0 increases , for fixed @xmath100 this inequality always occurs if @xmath199 , where @xmath200 is a threshold depending on @xmath98 , @xmath99 . the case @xmath201 , where @xmath202 is the critical threshold for the onset of the nekhoroshev regime , presents no practical interest . if , however , @xmath203 , then , for all values of @xmath0 in the interval @xmath204 the optimal normal form describes a true non - resonant dynamics . note that in order to describe the dynamics close to a point @xmath205 of the action space corresponding to diophantine frequencies @xmath206 , it suffices to choose @xmath51 such that @xmath207 corresponds to a very high order rational approximation of @xmath207 , i.e. the numbers @xmath208 are high order finite digit approximants of the numbers @xmath209 . then , @xmath100 becomes very large , and @xmath200 approaches very close to zero . in this case , for @xmath0 sufficiently small , we expect the existence of a set of points of large measure within @xmath85 , corresponding to kolmogorov - arnold - moser tori in the neighborhood of the point @xmath51 . however , these tori can not fill an open domain . thus , the diffusion in action space is topologically possible for ( very weakly ) chaotic orbits wandering through the set of kam tori . however , in the absence of significant resonant chaotic layers ( since no important resonances cross @xmath85 ) , the question of whether or not the diffusion can be observed is of no practical interest , since its rate would be extremely slow to be of any relevance in applications . thus , the non - resonant case is no further considered below . as long as @xmath210 , the point @xmath51 is doubly - resonant with respect to the optimal k truncation . in this case , the normal form contains either terms independent of the angles , or trigonometric terms of the form @xmath211 , @xmath212 and their multiples in the exponents . writing explicitly only the most important terms , the normalized hamiltonian takes the form : @xmath213 the main feature of the hamiltonian ( [ hamres ] ) is that , since in @xmath214 there are coupling terms between more than one resonant angles , the normal form @xmath26 _ alone _ is _ non - integrable_. in fact , @xmath26 can be decomposed into an integrable system of one degree of freedom and a non - integrable system of two degrees of freedom ( see @xcite ) . the decomposition is done by the linear canonical transformation @xmath215 @xmath216 @xmath217 defined by @xmath218 where @xmath101 has been defined in eq.([mvec ] ) . the hamiltonian in the new variables reads ( apart from a constant ) @xmath219 where @xmath220 and the remainder @xmath221 is exponentially small in @xmath29 . since @xmath222 is ignorable in @xmath26 , @xmath223 is an integral under the flow of the normal form . on the other hand , the remaining degrees of freedom @xmath224 and @xmath225 are coupled under the flow of @xmath26 due to the trigonometric terms @xmath226 . the main characteristics of motion can be understood by the following remarks : refer to non - scaled values , i.e. before the re - scaling of eq.([resc ] ) is implemented , while all action variables defined by a symbol starting with the letter @xmath139 have re - scaled values , according to eq.([resc ] ) . thus , in the domains considered below , all quantities of the form @xmath227 , where @xmath51 is the selected central doubly - resonant point of interest , scale proportionally to @xmath114 , while all actions denoted by a letter @xmath139 exhibit no scaling with @xmath0 . furthermore , all hamiltonian - type functions denoted by @xmath228 , @xmath162 , @xmath26 , or @xmath27 , are expressed in re - scaled variables ; only the original hamiltonian ( eq.([hamgen ] ) ) is expressed in non - scaled action variables @xmath229 . finally , the quantities @xmath230 ( eq.([eneprime ] ) ) and @xmath41 ( eq.([ez ] ) ) scale proportionally to @xmath114 . ] \i ) the constant - valued action @xmath223 can be viewed as a parameter in the two degrees of freedom hamiltonian @xmath26 . furthermore , except for the case of some very low resonances satisfying @xmath231 , all coefficients @xmath232 in ( [ nfdble ] ) are of order @xmath114 or higher . thus , the terms @xmath233 define an ` integrable part ' of @xmath26 , while the remaining terms depending on the resonant angles can be considered as a perturbation . the terms quadratic in @xmath234 in the r.h.s . of ( [ nfdble0 ] ) define the quadratic form @xmath235 in appendix a it is demonstrated that , due to the quasi - convexity condition assumed for the hessian matrix @xmath50 , the quadratic form ( [ z02 ] ) is positive definite . thus , the constant level curves of the quantity @xmath236 on the plane @xmath237 , given by @xmath238 are ellipses centered at @xmath239 ( the role of the elliptic structures formed around double resonances in the nekhoroshev theorem is discussed extensively in @xcite ) . if the higher order terms in the action variables of the development of eq.([nfdble0 ] ) are taken into account , the constant energy condition of eq.([eneprime ] ) yields deformed ellipses on the plane @xmath237 . if @xmath240 or @xmath241 , the slow frequencies @xmath242 , @xmath243 are non - zero , and they are given by @xmath244 on the other hand , due to the definition ( [ restra ] ) one has @xmath245 which is valid for any value of @xmath246 in the domain of convergence of the series ( [ nfdble0 ] ) . it follows that all the resonant manifolds defined by relations of the form @xmath247 intersect any of the planes @xmath237 corresponding to a fixed value of @xmath223 . using the notation @xmath248 the intersection of one resonant manifold with the plane @xmath237 is a curve . in the linear approximation , we have @xmath249 the above equation defines a ` resonant line ' , which is the local linear approximation to a ` resonant curve ' . all resonant lines ( or curves ) pass through the point @xmath250 , which , therefore , belongs to the resonant junction defined by the wavevectors @xmath80 . to each resonant curve we can associate a resonant strip in action space whose width is proportional to the separatrix width for that resonance . if , for a single pair of integers @xmath251 , we only isolate the resonant terms @xmath252 in the normal form @xmath26 ( eq.([hamres ] ) ) , we obtain a simplified resonant normal form @xmath253 corresponding to the limiting case of a single resonance . in a strict sense , @xmath254 describes well the dynamics far from the resonant junction . however , it can also be used in order to obtain estimates of the resonance width along the whole resonant curve defined by the integer pair @xmath251 . to this end , the leading terms of @xmath253 are ( apart from constants ) : @xmath255 + \ldots\nonumber\end{aligned}\ ] ] where the coefficients @xmath256 satisfy the estimate @xmath257 ) . after still another transformation @xmath258 , @xmath259 , @xmath260 , @xmath223 becomes a second integral of motion of @xmath253 , which takes the form @xmath261\end{aligned}\ ] ] where @xmath262 and @xmath263 are constants of the hamiltonian flow of ( [ hamresn12n ] ) . combining ( [ gn12 ] ) and ( [ hamresn12n ] ) , the separatrix width can be estimated as @xmath264 eq.([sepwidth ] ) allows to estimate the width of a resonant strip in the direction normal to a resonant curve on the plane @xmath237 . using the relations @xmath265 ( for @xmath266 ) , this estimate takes the form @xmath267 by the overlapping of various resonant strips whose limits ( pairs of parallel red lines ) correspond to separatrix - like thin chaotic domains around each resonance . two constant normal form energy ellipses @xmath268 and @xmath269 are also shown . right : the front and back panels show the phase portraits corresponding to a surface of section ( in one of the pairs @xmath270 or @xmath271 ) under the normal form dynamics alone , for the energies @xmath268 ( front panel ) and @xmath269 ( back panel ) . the blue curly arrows in both panels indicate the directions of a possible ` drift ' motion ( = slow change of the value of @xmath230 ) due to the influence of the remainder on dynamics . ] the outcome of the analysis so far can be visualized with the help of figure [ fgdbleresmodel ] ( schematic ) . the left panel shows the structure of a doubly - resonant domain in the plane of the resonant action variables @xmath237 . the two bold ellipses correspond to the constant energy condition for two different values of @xmath230 , namely @xmath268 and @xmath269 with @xmath272 . their common center is the point @xmath273 defined in eq.([jr0 ] ) . the three pairs of parallel red lines depict the borders of the separatrix - like thin chaotic layers of three resonances passing through the center . infinitely many such resonances exist , corresponding to different choices of integer vectors @xmath274 ; however , their width decreases as @xmath275 increases , according to eq.([reswidth ] ) . we thus show schematically only three resonances with a relatively low value of @xmath275 , named by the letters ` a ' , ` b ' and ` c ' . the blue curly curves indicate a slow drift undergone by the chaotic orbits along the resonance layers , allowing for a transition from one resonance to another . this phenomenon , which will be addressed in detail below , is due to the influence of the _ remainder terms _ of the normalized hamiltonian on dynamics . here , however , we discuss first the ( non - trivial ) influence of the _ normal form terms _ on dynamics , by considering the hamiltonian flow under the approximation @xmath276 . then , the following facts hold : + - for any fixed value of @xmath230 , and a fixed section in the angles , the motion is confined on one ellipse . + - for @xmath230 large enough ( @xmath268 , outermost ellipse in the left panel of fig.[fgdbleresmodel ] ) , the various resonant strips intersect the ellipse @xmath268 at well distinct arcs , i.e. there is no resonance overlap . the right front panel in fig.[fgdbleresmodel ] shows schematically the expected phase portrait , which can be obtained by evaluating an appropriate poincar surface of section , e.g. in the variables @xmath224 or @xmath225 . the dashed lines show the correspondence between the limits of various resonant domains depicted in the left and right panels . in particular , the intersection of each resonant strip in the left panel with the ellipse @xmath268 corresponds to the appearance of an associated _ island chain _ in the right panel . the size of islands is given essentially by the separatrix width estimate of eq.([reswidth ] ) . hence , the size of the islands decreases exponentially with the order of the resonance @xmath277 . however , the main effect to note is that , since all resonant strips are well separated on the ellipse , the thin separatrix - like chaotic layers marking the borders of each of their respective island chains do not overlap . as a result the local chaos around one resonance is isolated from the local chaos around the other resonances . in fact , the normal form dynamics induces the presence of rotational kam tori which , in this approximation ( @xmath276 ) , completely obstruct the communication among the resonances . note that a detailed study of the dynamics of the above type , induced by the doubly - resonant normal form , was recently presented in @xcite . + - far from the domain of resonance overlap , the size of the islands corresponding to each resonance is nearly independent of the energy @xmath230 , as it depends essentially only on the size of the fourier coefficient of the corresponding harmonics in the hamiltonian . however , the separation of the islands is reduced as the energy _ decreases _ , since this separation is given essentially by the separation between the distinct arcs in fig.[fgdbleresmodel ] at which the various resonances intersect the ellipse corresponding to a fixed energy @xmath230 . as a result , below a critical energy @xmath278 , significant resonance overlap takes place , leading to the communication of the chaotic layers of the various resonances and an overall increase of chaos . this is shown in the left panel of fig.[fgdbleresmodel ] for an ellipse @xmath279 , with the corresponding phase portrait shown in the right back panel . we note in particular the ` merging ' of all three resonant domains one into the other , which produces a large connected chaotic domain surrounding all three island chains ( and many other smaller chains , not visible in this scale ) . the value of the critical energy @xmath280 marking the onset of large scale resonance overlap can be estimated as follows : each resonant strip intersects one fixed energy ellipse on one arc segment . also , eq.([reswidth ] ) can be replaced by the estimate @xmath281 where @xmath277 , @xmath282 , and @xmath283 , with the constants @xmath284 defined as in eq.([mh ] ) . the total length @xmath285 of all segments can be now estimated by summing , for all @xmath13 , the estimate ( [ reswdk ] ) , namely @xmath286 on the other hand , the total circumference of the ellipse for the energy @xmath230 is estimated as @xmath287 where @xmath288 is the geometric mean of the ellipse s major and minor semi - axes . for @xmath288 one has the obvious estimate @xmath289 , whence @xmath290 the critical energy @xmath291 can now be estimated as the value where @xmath292 , implying that the associated ellipse is fully covered by segments of resonant strips . thus @xmath293 eq.([epcrit ] ) implies that @xmath280 is a @xmath294 quantity . when @xmath309 , @xmath51 is simply - resonant with respect to the optimal k truncation . in this case , the normal form contains terms either independent of the angles , or depending on them via trigonometric terms of the form @xmath310 , @xmath311 . using the same notations as in the previous subsection , the transformed hamiltonian reads : @xmath312 , but for a simple resonance . in this case , any other resonance crossing the main ( guiding ) resonance has an exponentially small width and acts as a ` driving ' resonance for diffusion . ] repeating all steps as in the case of double resonance leads to the normal form @xmath313 the main difference with respect to the doubly - resonant normal form ( [ hamresn12 ] ) is that , the angle @xmath314 being ignorable , the action @xmath315 ( or @xmath316 ) is an integral of the flow of @xmath254 , in addition to @xmath223 . thus , @xmath254 defines an integrable hamiltonian . a pair of constant values @xmath317 , @xmath318 defines a straight line @xmath319 which corresponds to the unique resonance @xmath320 . this will be called ` main resonance ' (= the ` guiding resonance ' in @xcite ) . in figure [ fgspleresmodel ] ( schematic ) , the domain of the main resonance is delimited by two vertical thick red lines corresponding to the separatrix - like thin chaotic layers at the boundary of the resonance similarly to fig.[fgdbleresmodel ] . using similar arguments as in the derivation of eq.([reswdk ] ) , the separatrix width can be estimated as @xmath321 under the normal form dynamics , motions are allowed only across the resonance , i.e. in the direction @xmath322 . in fig.[fgspleresmodel ] this is the horizontal direction . the thin strip delimited by two horizontal red lines corresponds to the resonance with resonant wavevector @xmath99 , which , since @xmath323 , is now of width exponentially small ( @xmath324 ) . thus , it will be called a ` secondary ' resonance . in order to estimate the speed of diffusion as a function of the optimal remainder in this case , let us note first that the influence of the remainder on dynamics is to slowly change the value of the two approximate integrals @xmath223 and @xmath316 , that would be exactly preserved under the normal form dynamics . in view of eq.([hamresn1 ] ) , the hamiltonian ( [ hamressple0 ] ) can be approximated by @xmath325+ ... \bigg]\end{aligned}\ ] ] where i ) the ( non - integer ) vectors @xmath326 , @xmath79 come from the solution of the right eqs.([restra ] ) for the angles @xmath327 in terms of the angles @xmath328 , @xmath314 , and @xmath329 , and ii ) we approximate all the fourier coefficients in the remainder series by their constant values @xmath330 at the points @xmath331 ( we set @xmath332 for @xmath333 ) . the latter approximation is sufficient for estimates regarding the speed of diffusion . the key remark is that for all the coefficients @xmath330 the bound @xmath334 holds , while , for the leading fourier term @xmath335 in the remainder we have @xmath336 . in fact , we typically find that the size of the leading term is larger from the size of the remaining terms by several orders of magnitude , since this term contains a repeated product of small divisors of the form @xmath337 ( see appendix a ) . furthermore , using an analysis as in @xcite , we readily find @xmath338 , where @xmath339 is a so - called ( in @xcite ) ` delay ' constant . we note in passing that the fourier terms of the form @xmath335 are called ` resonant ' in @xcite . the value of the diffusion coefficient can now be estimated by applying the heuristic theory of chirikov ( @xcite , see also @xcite and @xcite ) in the hamiltonian model ( [ hamressple ] ) . the estimate @xmath340 holds , where @xmath341 , @xmath342 is an average period of motion within the main resonance separatrix - like thin chaotic layer , of width @xmath343 , @xmath61 is the melnikov function with argument @xmath344 ( see appendix b of @xcite ) , the vector @xmath345 being defined by the relation @xmath346 . the estimate @xmath347 holds . in view of eq.([restra ] ) however , we have that @xmath348 . since @xmath349 ( see appendix b ) , and @xmath350 , it follows that @xmath351 , for an exponent @xmath31 . putting these estimates together , we finally arrive at a steeper dependence of the diffusion coefficient @xmath1 on the optimal remainder @xmath2 in the case of simple resonance than in the case of double resonance , namely : @xmath352 regarding now the precise value of @xmath4 , it is hardly tractable to determine this on the basis exclusively of the behavior of the melnikov integrals discussed above . we note , however , that the quantity @xmath353 yields the size of the ` splitting'@xmath21 of the separatrix of the main ( guiding ) resonance due to the effects of the leading term in the remainder function . the relation between the separatrix splitting and the size of the optimal remainder has been examined in @xcite and later in @xcite . in the latter work , the estimate @xmath354 was predicted and probed by numerical experiments , where @xmath138 ( in the notation of @xcite ) is the effective size of the perturbation to the normal form pendulum dynamics caused by the remainder . setting thus @xmath355 suggests the scaling @xmath356 , whereby the constant @xmath4 can be estimated as @xmath357 . hence ( in view of [ difcfsple ] ) @xmath358 in simply resonant domains . despite the heuristic character of the above derivation , it seems that the value @xmath357 is supported by the results of numerical experiments . in particular , in @xcite the diffusion coefficient @xmath1 along a simple resonance was compared directly to the size of the optimal normal form remainder . it was found that @xmath37 , essentially confirming that @xmath359 . we point out , however , that in @xcite a different exponent was found @xmath360 regarding the same resonance as in @xcite , while it was found that @xmath23 in the case of a very low order simple resonance ( with @xmath231 ) , which is not discussed in our present work . these exponents , on the other hand , depend on the chosen definition of the numerical measure used to estimate both @xmath21 and @xmath2 . thus , a detailed quantitative comparison of the works cited above is left as on open problem for future study . in our numerical work we employ the same hamiltonian model of three degrees of freedom as in @xcite . the hamiltonian reads : @xmath361 this model has a particularly simple , yet sufficient for our purpose , structure , allowing to probe numerically all steps of the previous section . in particular : the function ( [ hamfr ] ) is polynomial in the action variables , thus it is analytic in any complex extension of @xmath362 . on the other hand , the domain of analyticity in the angle variables was examined in @xcite . it was found that analyticity can be established in a set @xmath363 , @xmath364 , @xmath79 , for a positive constant @xmath52 estimated semi - analytically as @xmath365 . accordingly , the coefficients @xmath63 of the fourier development @xmath366 where @xmath75 , @xmath367 , decay exponentially . the distance of the nearest singularity , with respect to each of the angles @xmath368 , from the real axis is given by the solution of @xmath369 , or @xmath370 . thus , the following bound holds : latexmath:[\[\label{expdecay } as regards convexity , for all @xmath67 the matrix @xmath66 has a particularly simple structure , since we have @xmath372 , and @xmath373 for all other @xmath374 . thus there are two positive eigenvalues equal to unity and one equal to zero , while @xmath375 . the constant energy condition @xmath376 defines a paraboloid in the action space . the resonant manifolds are planes , since @xmath377 , @xmath378 , @xmath379 , whereby the resonance conditions @xmath380 for all @xmath75 define planes normal to the @xmath381 plane . it follows that , when projected to the @xmath381 plane , the intersections of all resonant manifolds with a surface of constant energy of the unperturbed problem yield a set of straight lines . this greatly facilitates the numerical study , since all diffusing orbits in the perturbed problem follow piecewise straight paths nearly parallel to one or more resonant lines of the unperturbed problem , while the orbits can only change direction by approaching close to resonance junctions . examples of diffusion of this type along a simple resonance where studied in @xcite , while the case of consecutive encounters with doubly - resonant domains was examined in a mapping model @xcite variant of the hamiltonian model ( [ hamfr ] ) . the connection between the size of the optimal remainder @xmath2 and the diffusion coefficient @xmath1 in a case of simple resonance was the main subject of a previous study @xcite . following the same terminology and notations as in section 2 above , the point @xmath51 in the normal form construction in @xcite was chosen as @xmath382 . for this point we have ( viz . eq.([mvec ] ) ) @xmath383 , @xmath384 , @xmath385 . the optimal truncation order in all calculations of @xcite varied from @xmath386 to @xmath387 ( depending on the value of @xmath0 in the range considered ) . thus , in all cases we have @xmath388 , that is the so - chosen point @xmath51 was found to be simply resonant with respect to the optimal k truncation . following fig.5 of @xcite it was then found by numerical fitting that the diffusion coefficient @xmath1 scales with the optimal remainder as @xmath37 . a theoretical justification for this ` steepening ' of the power - law with respect to the exponent @xmath389 holding in double resonances was given in subsection 2.3.2 . in order to probe now the dependence of @xmath1 on @xmath2 in the case of a double resonance , in the sequel we focus our numerical study on a different point of @xmath88 , namely @xmath390 . the basic resonant wavevectors are @xmath391 the hamiltonian normalization is carried out as exposed in subsection 2.2 . the interval of values of @xmath0 considered is @xmath392 which , according to @xcite is below the critical value for the onset of the ` nekhoroshev regime ' ( @xmath393 ) . furthermore , it will be shown below that for all values of @xmath0 in the above interval the optimal fourier - truncation order @xmath198 turns to be much larger than @xmath394 . on the other hand , for the basic wavevectors we have @xmath395 , @xmath396 . thus , for all considered values of @xmath0 one has @xmath397 , that is , the point @xmath51 is doubly resonant with respect to any of the optimal k truncations considered in the sequel . due to eq.([kprime ] ) , the constant @xmath121 in terms of which book - keeping is implemented changes with @xmath0 . however , one notices that , because of the logarithmic dependence of @xmath121 on @xmath0 , in the largest part of the interval @xmath392 , where we focus , one has a constant value @xmath398 , while one has @xmath399 only close to the upper limit @xmath400 and @xmath401 close to the lower limit @xmath402 . for simplicity , we thus fixed the value of @xmath121 as @xmath398 in all normal form computations . doing so , computer memory limitations restrict all computed expansions to a maximum order @xmath403 in the book - keeping parameter @xmath127 , or maximum order @xmath404 in fourier space . in fact , for @xmath405 we perform at most 14 normalization steps , so that the remainder contains terms of at least three consecutive orders in @xmath127 , namely @xmath406 , @xmath407 and @xmath408 . as explained below , this allows us to perform some numerical tests regarding the convergence of the remainder series when the optimal normalization order is as high as @xmath158=14 ( or @xmath409 ) . on the other hand , for @xmath410 we allow for one more normalization step ( @xmath406 ) in order to get as close as possible to the optimal order , which , as shown below for @xmath411 is larger than 14 . thus , for the calculation of the corresponding remainder value at this order ( @xmath406 ) we necessarily have to rely on the sum of only two rather than three or more consecutive terms . writing the truncated ( at order 17 ) remainder function as : @xmath412 where @xmath413 are the terms of order @xmath414 in the book - keeping parameter @xmath127 allows us to probe numerically the convergence of the remainder function within any chosen domain @xmath85 in action space . to this end , at any normalization order @xmath148 , let us consider a disk @xmath415 in the space of the transformed action variables ( we neglect the action @xmath416 which , in the particular case of the hamiltonian ( [ hamfr ] ) , is dummy , i.e. it does not appear in any higher order term of either the normal form or the remainder ) . this is a deformed disk also in the old canonical variables @xmath417 , limited by a boundary given approximately by @xmath418 ( cf . the action re - scaling given by eq.([resc ] ) ) . for the hamiltonian ( [ hamfr ] ) one can readily check that all the terms in @xmath413 are trigonometric polynomials of maximum degree @xmath419 whose coefficients are polynomial of maximum degree @xmath420 in the actions , namely @xmath421 furthermore , in the disk @xmath422 the obvious bound @xmath423 holds . we thus define the norm @xmath424 can be obtained . in fact , by calculating the truncated sums @xmath425 for any fixed choice of @xmath43 , where @xmath426 takes all values @xmath427 , we can have a clear numerical indication of whether the remainder function was calculated up to a sufficiently high order for convergence to have been practically reached . the maximum value of @xmath43 for which the series @xmath428 converges absolutely sets the size of the doubly - resonant domain @xmath429 ( in non - scaled variables ) where the normal form calculations are valid . in practice , we are interested in the diffusion of orbits with initial conditions inside this domain . in particular , in subsection 3.2 we will consider orbits starting on the circle @xmath430 . all our numerical orbits are studied up to a time in which their distance from the center of the double resonance changes significantly less than @xmath431 ( see below ) . variations of this order at maximum are found when we measure @xmath43 either in the original canonical action variables or in the variables after the optimal canonical transformation . thus , for all the orbits we can set a safe outer boundary @xmath432 within which they are well confined . we then verify numerically that this domain belongs to the analyticity domain of the various transformations employed in the form of series ( of the new variables in terms of the old variables or vice versa ) . this check is made by finding whether the fourier coefficients of the series exhibit an exponential decay . an example is given in fig.[fggksigma ] . we consider the fourier series yielding the new transformed canonical action @xmath433 as a function of the old canonical variables , for @xmath434 , at the normalization orders @xmath435 and @xmath436 . writing this as a series @xmath437 we define the coefficients @xmath438 ( see text ) as a function of the fourier order @xmath439 , for @xmath434 , at the normalization orders @xmath440 , @xmath441 , and @xmath442 ( lower , middle and upper set of points respectively ) . all three curves exhibit an exponential decay for large @xmath439 , with nearly the same asymptotic law . the straight line has inclination @xmath443 . ] figure [ fggksigma ] shows the coefficients @xmath444 for @xmath434 , @xmath445 , and @xmath435 and @xmath436 . we observe that all three curves exhibit a tail showing exponential decay of the fourier coefficients . however , it is remarkable that the asymptotic exponential slope seems to change only marginally . instead , the main change , as @xmath148 increases , regards that formation of a ` plateau ' of fourier coefficients of nearly constant size formed for small @xmath439 . namely , the width of the plateau increases as @xmath148 increases . it is remarkable that the asymptotic tail laws for all @xmath148 appear to follow an exponential decay with the same constant @xmath446 , i.e. with nearly the same value as the constant appearing in the analyticity condition of the original hamiltonian ( cf . eq.([expdecay ] ) ) . this effect shows that , while in the usual proofs of the nekhoroshev theorem one requires a reduction of the analyticity domain at every normalization step , i.e. one considers bounds of the form @xmath447 with @xmath448 , in practice the dependence of the coefficients @xmath449 on @xmath439 is more complicated than a simple exponential decay law . in fact , the constants @xmath450 reflect an average exponential slope that compensates between the plateau , for small @xmath439 , and the exponential tail , for large @xmath439 . namely , as the width of the plateau increases with @xmath148 , one obtains smaller and smaller values of the average exponential decay constant @xmath450 . as a function of the truncation order @xmath426 when @xmath434 , @xmath430 , and the normalization orders are @xmath451 ( upper curve ) , @xmath442 ( lower curve ) and @xmath452 ( middle curve ) . ( b ) the value of @xmath453 as a function of @xmath148 for different values of @xmath0 . for @xmath454 and @xmath455 , the dashed curves after the order @xmath406 are found by quadratic extrapolation . no attempt to extrapolate was made for @xmath456 and @xmath402 . ( c ) the optimal normalization order @xmath158 as a function of @xmath0 together with a power - law best fitting curve . ] fig.[fgremopt]a shows now an example of the behavior of the truncated remainder function for @xmath457 and @xmath434 . the upper curve shows the value of @xmath458 at the normalization order @xmath451 as a function of @xmath426 for @xmath459 . clearly , after @xmath460 the cumulative sum ( [ sumrem ] ) shows no further substantial variation , which indicates that the remainder series converges after three consecutive terms @xmath461 and @xmath462 ( this is verified also by computing numerically a convergence criterion like dalembert s criterion ) . the lower and middle curves show now the same effect for the normalization orders @xmath442 and @xmath452 respectively . note that the three consecutive truncation orders @xmath463 and @xmath408 allowed for the computation of the remainder at the normalization order @xmath452 are essentially sufficient to demonstrate the convergence of the remainder . hence , @xmath464 represents a good numerical estimator of the value of the remainder series for normalization orders up to @xmath452 . however , the main effect to note is that the estimated remainder value @xmath465 found for @xmath452 is larger than the one for @xmath442 , implying that the _ optimal _ normalization order @xmath158 is below @xmath452 . fig.[fgremopt]b shows , precisely , the asymptotic character of the above normalization , showing @xmath465 against the normalization order @xmath148 for various values of @xmath0 as indicated in the figure . for all values down to @xmath466 we now observe the asymptotic behavior , namely the size of the remainder initially decreases as @xmath148 increases , giving the impression that the normalization might be a convergent procedure . however , this trend is reversed after an optimal order @xmath158 , where the remainder reaches its minimum value , while , for @xmath467 the remainder increases with @xmath148 and eventually goes to infinity . we also observe that for @xmath468 the optimal order is beyond @xmath406 . however , for @xmath454 and @xmath455 , the computed remainder values are close to the minimum . the dashed extensions of the numerical curves shown in fig.[fgremopt]b correspond to an extrapolation obtained by quadratic fitting of the available numerical points near the corresponding minima . using this extrapolation , we obtain an estimate of the optimal remainder size for the values @xmath454 and @xmath455 , that will be used in some calculations below . on the other hand , for @xmath456 and @xmath402 , even using the extrapolation we find that the optimal normalization is beyond any reliable possibility to estimate given our computing limitations . as discussed above , the estimate @xmath469 holds @xcite , i.e. @xmath158 is expected to be a decreasing function of @xmath0 . fig.[fgremopt]c shows the numerical estimate for @xmath158 as a function of @xmath0 from the points of minima of fig.[fgremopt]b . the blue curve is a power - law fitting , yielding the exponent 0.52 , i.e. very close to the one predicted by theory . since the value @xmath470 depends on @xmath0 , from the above procedure we obtain numerically pairs of values @xmath471 . in subsection 3.4 below , we will numerically calculate the value of the diffusion coefficient @xmath1 for each one of the selected values of @xmath0 , thus allowing for a probe of the dependence of @xmath1 on @xmath2 and a comparison with the results of subsection 2.3.1 . the resonant structure in the action space ( around @xmath51 ) can be visualized by employing the method of the fli map as in @xcite . we recall that the fast lyapunov indicator ( fli ) is a numerical indicator of chaos , defined for one orbit by @xmath472 where @xmath473 is a deviation vector , i.e. in our case @xmath474 found after solving the variational equations of motion up to the time @xmath16 from some initial conditions @xmath475 . by properly choosing a threshold value @xmath476 , orbits with @xmath477 are characterized as regular , and those with @xmath478 as chaotic . furthermore , a convenient use of the fli in the visualization of the arnold web is found by producing fli color maps @xcite . considering a grid of initial conditions in the action space , we assign to each initial condition a color corresponding to the fli value found for the resulting orbit integrated up to a sufficiently long time ( of the order 100 1000 periods ) . this allows for illustrating the resonant structure in action space , as shown in fig.[fgdbleres ] , which is an fli map in an action domain including our chosen doubly - resonant point @xmath479 for three different values of @xmath0 . in all three panels , there are resonances projecting on @xmath381 as single yellow or orange thick lines , while other resonances project as strips with a green or blue interior zone delimited by pairs of nearly parallel yellow or red lines . as explained in @xcite , this difference is only due to the particular choice of surface of section ( @xmath480 , @xmath481 , similar to @xcite ) . namely , the yellow lines marking all resonances represent the intersection of the thin separatrix - like chaotic layers formed around each resonance with the chosen surface of section . this produces a pair of nearly parallel yellow or orange lines for any resonance ( of the form @xmath482 ) whose leading fourier coefficient @xmath63 of the resonant term @xmath113 in the original hamiltonian expansion has a negative real part , while it produces a single yellow or orange thick line if @xmath483 is positive . in the latter case , the domain of regular orbits inside the resonance has no projection on the chosen surface of section , while in the former case it projects as a strip of green or blue color . of the hamiltonian [ hamfr ] for @xmath480 , @xmath484 , around the doubly - resonant point @xmath479 for ( a ) @xmath402 , ( b ) @xmath466 , ( c ) @xmath485 . the color scale represents the computed value of the fli ( see text ) in the intervals @xmath486 ( magenta , most ordered ) , @xmath487 ( blue ) , @xmath488 ( green ) , @xmath489 ( orange ) , @xmath490 ( yellow , most chaotic ) . ] when @xmath402 ( fig.[fgdbleres]a ) , we easily distinguish four main resonances passing through @xmath479 . the biggest resonant domain ( green , from bottom left to top right ) corresponds to the resonance @xmath491 , whose corresponding wave - vector is the basic resonant wavevector @xmath98 . similarly , the single yellow - red thick line going from bottom right to top left is the resonance @xmath492 , whose corresponding ( also basic ) wavevector is @xmath99 . we also clearly distinguish two resonances of order @xmath493 , namely @xmath494 ( blue ) , and @xmath495 ( green ) . many other higher order resonances cross the central doubly - resonant point @xmath479 , denoted hereafter by o , but they are not so visible in the scale of fig.[fgdbleres]a . the resonant strips of all previous resonances join each other forming a domain of double resonance around o. the extent of this domain can be determined roughly by drawing concentric circles around the point o. such circles correspond to nearly constant normal form energy values , as can be seen by noting that , for the particular hamiltonian function ( [ hamfr ] ) , the coefficients @xmath496 of eq.([hamresn12 ] ) have the values @xmath497 @xmath498 , and @xmath499 . applying eqs.([z02],[ene0],[jr0 ] ) for the particular resonant wavevectors given by ( [ resvectors ] ) , the doubly - resonant normal form of the hamiltonian ( [ hamfr ] ) expressed in resonant variables takes the form @xmath500 where the @xmath501 terms are trigonometric polynomials of the resonant angles @xmath502 , @xmath503 , while @xmath504 is a constant which appears only in the numerical values of the quantity @xmath505 called , hereafter , the normal form energy ( @xmath41 differs from the quantity @xmath230 defined in eq.([eneprime ] ) only by the constant @xmath506 ) . we note that the estimate @xmath501 for the trigonometric terms in @xmath26 follows from the estimate ( [ gn12 ] ) for the size of the corresponding fourier coefficients , taking into account that @xmath507 , according to eq.([kprime ] ) . since the angle @xmath222 is ignorable in the hamiltonian ( [ ene02fr ] ) , @xmath223 is an integral of the flow of @xmath26 . furthermore , since for the particular choice of hamiltonian model ( [ hamfr ] ) the action @xmath416 is dummy , implying that @xmath416 can be assigned any arbitrary value without affecting the dynamical evolution of any other canonical variable , we can always choose the value of @xmath416 so that @xmath508 . then , the normal form energy condition @xmath509 implies @xmath510 where @xmath43 is a @xmath511 quantity . transforming to the original non - scaled action variables we also find @xmath512 whereby @xmath513 is interpreted as the radius of a circle , around o , corresponding to a constant normal form energy condition . it follows that the set of all possible normal form energy values are represented on the @xmath381 plane as a set of concentric circles around o. three such circles are drawn in fig.[fgdbleres]a , corresponding to @xmath402 and @xmath514 ( outer circle ) , @xmath515 ( middle circle ) , and @xmath516 ( inner circle ) . their main difference concerns the degree of resonance overlapping in each case . namely , for a value of @xmath517 , corresponding to the outer circle @xmath514 , the main visible resonances of fig.([fgdbleres]a ) intersect the circle in some arcs only , while the remaining parts of the circle lie in the regular ( non - resonant ) domain . in the latter parts , the normal form dynamics alone would imply the existence of a set of kolmogorov - arnold - moser invariant tori of large measure . on the contrary , in the inner circle , corresponding to @xmath518 , all resonances essentially overlap , producing a strongly chaotic domain . the middle circle corresponds to @xmath519 , which is close to the critical energy below which resonance overlapping dominates the dynamics . the remaining panels of fig.[fgdbleres ] show what happens when @xmath0 is increased by a factor 5 ( @xmath466 , fig . [ fgdbleres]b ) , or 15 ( @xmath485 , fig . [ fgdbleres]c ) with respect to fig.[fgdbleres]a . a main feature to notice is that , by increasing @xmath0 , many more resonances ` show up ' in the fli map . furthermore , the size of all resonant domains grows proportionally to @xmath114 , as verified in fig.[fgdbleres ] , where by augmenting the scale in panels ( b ) and ( c ) by a factor @xmath520 and @xmath521 respectively with respect to panel ( a ) , the widths of all resonant strips passing through @xmath522 remain essentially unaltered in all three panels . thus , the only essential change is the increase of chaos as @xmath0 increases . namely , we see that the chaotic layers delimiting the borders of each resonance become thicker as @xmath0 increases . this also increases the resonance overlapping locally , close to the points of resonance crossings . defined as in eqs.([restra ] ) for the hamiltonian ( [ hamfr ] ) , for the same surface of section and using the same color scale as in fig.([fgdbleres ] ) . the circle in ( a ) corresponds to the constant normal form energy value @xmath523 in ( a ) and @xmath524 in ( b ) . the phase portraits of the normal form dynamics for the values ( c ) @xmath523 and ( d ) @xmath525 . the plotted surfaces of sections are @xmath270 whenever the quantity @xmath526 ( where all symbols denote the new canonical coordinates and momenta after the optimal lie normalization ) crosses a multiple value of @xmath527 . the main resonances are identified as : @xmath491 ( vertical ) , @xmath492 ( horizontal ) , @xmath494 ( bottom left to top right diagonal ) , @xmath495 ( top left to bottom right diagonal ) . ] focusing , now , on one value @xmath528 , figure [ fgresportrait ] shows in detail the implications of normal form dynamics in the two regimes when there is no resonance overlap ( @xmath529 , figs . [ fgresportrait]a , c ) , or when there is substantial resonance overlap ( @xmath524 , figs.[fgresportrait]b , d ) . the upper panels correspond to fli maps as in fig.([fgdbleres ] ) . here , however , instead of the action variables ( @xmath530 ) we use the resonant re - scaled actions @xmath237 , defined as in eq.([restra ] ) , where , for each point in the action space of the original variables we compute the values of the transformed actions @xmath531 , @xmath79 by the composition of the lie canonical transformations resulting from the computer - algebraic program calculating the optimal normal form @xmath26 . since the same program renders also the algebraic form of @xmath26 , we use this expression to derive the hamiltonian equations of motion of the normal form alone , namely @xmath532 , @xmath533 , @xmath534 , @xmath535 , @xmath536 , while we set @xmath537 . for each value of @xmath41 , we then compute _ numerical orbits under the normal form dynamics _ alone via the previous equations . finally we plot a convenient surface of section of the normal form flow , taken by the condition @xmath538 . these sections are shown in figs.[fgresportrait]c , d , for the normal form energy values @xmath529 and @xmath539 respectively . the corresponding circles , through eq.([eres ] ) , are shown in panels ( a ) and ( b ) , superposed to the color background yielding the fli map in the resonant action variables for @xmath528 . the main feature of this plot is the exact correspondence between the values of @xmath540 where each resonance intersects the circle corresponding to @xmath509 in panels ( a ) and ( c ) , and the projection of these values to thin chaotic layers delimiting the same resonance in the corresponding surface of section . in fact , inside each resonance we have regular orbits corresponding to islands of stability on the surface of section . furthermore , while at the normal form energy value @xmath529 there are many rotational kam tori separating these resonances , at the value @xmath524 these tori are destroyed and substantial resonance overlap takes place . this fact leads to the creation of a connected chaotic domain surrounding all main resonances in the surface of section of fig.[fgresportrait]d . this , in turn , implies that under the normal form dynamics alone no communication is allowed from one resonance to the other for the normal form energy value @xmath529 ( which in this approximation remains constant in time ) , while such communication is possible throughout the whole connected chaotic domain for @xmath524 . in fact , the phase portrait of fig.[fgresportrait]d renders visually clear that chaos is rather strong in this case . however , as emphasized in section 2 , this fact has no consequences regarding the possibility of long excursions in the action space , since all motions in this approximation would be bounded on circles like those of figs.[fgresportrait]a , b . on the contrary , such excursions are only possible due to the effect of the remainder , which causes the chaotic orbits to slowly ` drift ' from circle to circle as the value of @xmath41 changes slowly in time . to this we now turn our attention . the main effect , of local diffusion within the doubly - resonant domain , can now be demonstrated with the help of figure [ fgarnolddif ] . the time evolution of one chaotic orbit is shown in this figure , as the orbit moves within the doubly - resonant domain along some of the main intersecting resonances . in this example as well we take @xmath528 ( as in fig.[fgresportrait ] ) . ) for @xmath528 . after computing the optimal normal form , we find , via the lie canonical transformations , the values of all transformed variables @xmath541 and @xmath542 corresponding to particular values of the old variables @xmath543 and @xmath544 stored at many different times @xmath16 within an interval @xmath545 along the numerical run . using the numerical values of the computed transformed variables , ( a ) shows the variation of the normal form energy @xmath546 as a function of @xmath16 in the intervals @xmath547 ( blue ) , @xmath548 ( red ) , and @xmath549 ( green ) . the initial and final values are equal to @xmath550 , while the minimum value , occurring around @xmath551 is @xmath524 . ( b ) the evolution of the orbit in the action space @xmath237 , using the same colors as in ( a ) for the corresponding time intervals . in the first time interval ( blue ) , the orbit wanders in the thin chaotic layer of the resonance @xmath494 . in the second time interval ( red ) it jumps first to the resonance @xmath495 , and then to the resonance @xmath491 . in the third time interval ( green ) the orbit recedes from the doubly - resonant domain along the resonance @xmath491 . ( c ) 3d plot in the variables @xmath552 , visualizing arnold diffusion for the same orbit . taking 20 equidistant values of @xmath553 , @xmath554 in the interval @xmath555 , we first find the times @xmath556 in the interval @xmath557 when the normal form energy value @xmath546 of the numerical orbit approaches closest to the values @xmath553 . for each @xmath558 , starting with the momentary values of all resonant variables at @xmath556 , we then compute 1000 poincar consequents of the normal form flow on the same section as in figs.[fgresportrait]c , d . the same procedure is repeated in a second interval @xmath559 . as a net result , the orbit at the beginning and end of the calculation is found on the same section ( corresponding to @xmath529 ) , but in a different resonant layer , having by - passed the barriers ( invariant tori of the normal form dynamics ) via a third dimension ( here parameterized by the time - varying value of @xmath41 ) . ] in fig.[fgarnolddif ] , the evolution of the orbit is shown for a total time @xmath560 . the optimal normal form for @xmath528 has also been computed , whose optimal normalization order is @xmath561 , corresponding to an optimal fourier order @xmath562 . since the corresponding lie generating functions are known , we compute , via the composition of lie canonical transformations , the values of all transformed variables @xmath541 and @xmath542 corresponding to particular values of the old variables @xmath543 and @xmath563 stored at many different times during the numerical run , i.e. as @xmath16 varies within the interval @xmath545 . finally , since the exact algebraic expression for the normal form @xmath26 is known , we compute the precise numerical value of the normal form energy @xmath546 at the same times . fig.[fgarnolddif]a shows the variation of the normal form energy @xmath546 as a function of the time @xmath16 in the intervals @xmath564 ( blue ) , @xmath548 ( red ) , and @xmath549 ( green ) . the final time is such that the initial and final values of @xmath41 are equal , namely @xmath550 . on the other hand , as @xmath41 slowly changes during the run , it acquires a minimum value around @xmath551 , which is @xmath565 . such evolution corresponds to the process described schematically in fig.[fgdbleresmodel ] ( section 2 ) . namely , from the previous figure ( fig.[fgresportrait ] ) we conclude that the two extreme values of @xmath41 acquired during the numerical run are such that @xmath566 while @xmath567 , where @xmath568 is the critical energy corresponding to a large scale overlapping of resonances ( subsection 2.3.1 ) . furthermore , as we will see in the next subsection , the chaotic excursions of the orbits , and , consequently , time evolution of @xmath41 , can be approximated by a normal diffusion process . furthermore , the fastest evolution takes place in the intervals @xmath569 , and @xmath570 , in both of which the total variation of @xmath41 is of the order of @xmath571 , or a ` per step ' variation of the order of @xmath572 . it should be stressed that these extremely small variations are possible to unravel numerically only because we use the new canonical variables deduced by the normalizing sequence of lie canonical transformations . when the old variables are used , instead , we find that the there are large variations ( of order @xmath114 ) of all quantities depending on the actions . these variations are , in fact , dominated by the so - called ( in the nekhoroshev theory ) ` deformation ' effects ( which are also of order @xmath114 ) , hence completely covering the drift effects which are much smaller in size . this feature of the optimal canonical transformations will be exploited in the measurement of the diffusion coefficient @xmath1 as described in the next subsection . fig.[fgarnolddif]b shows the diffusion of the orbit in the action space @xmath237 , using the same colors as in fig.[fgarnolddif]a for the corresponding time intervals ( the background produced by the fli map is shown here in gray scale ) . in the first time interval ( blue ) , the orbit wanders chaotically within the thin chaotic layer of the resonance @xmath494 . it should be stressed that this wandering has a random walk character , i.e. the orbit makes several reversals of its drift direction , sometimes approaching and other times receding from the center of the double - resonance . on average , however , the drift is in the inward direction ( this is a statistical effect ; for other initial conditions the average drift turns to be outwards ) . in the second time interval ( red ) , the orbit jumps first to the domain of the resonance @xmath495 . now , however , the chaotic motion takes place with a relatively high speed ( of order @xmath114 ) in the direction across resonances . as a result , the orbit fills nearly ergodically the whole connected chaotic domain surrounding the main overlapping resonances , while , at the end of this time interval , the orbit is closer to the resonance @xmath491 . finally , in the third time interval ( green ) the orbit recedes from the doubly - resonant domain ( this is also a statistical effect ) being trapped along the domain of the resonance @xmath491 . in this way , at the time @xmath560 , the orbit is found at about the same distance from the center as initially ( at @xmath573 ) , but on a different resonance . fig.[fgarnolddif]c , now , shows a 3d plot in the variables @xmath552 , visualizing the ` third dimension ' along which the arnold diffusion progresses for the same orbit . from this plot we can clearly see the effect of the remainder , which can be considered as a very slow modification of the normal form dynamics acting on a timescale of the order of @xmath574 periods . the normal form dynamics , on the other hand , describes well the motion over shorter timescales , of the order of @xmath575@xmath576 periods . in order to show the dynamical effects happening on both timescales , we adopt the following numerical procedure : taking 20 equidistant values of @xmath553 , @xmath554 in the interval @xmath555 , we first find the times @xmath556 within the interval @xmath557 ( where the motion is , in general , in the inward direction ) when the normal form energy value @xmath546 of the numerical orbit approaches the closest possible to the values @xmath553 . then , for each @xmath558 , we set the momentary values of all canonical variables of the numerical orbit at the time @xmath556 as initial conditions via which we compute the corresponding values of all the new resonant canonical variables following the composition of the corresponding lie canonical transformations . with these values as initial conditions , we compute 1000 poincar consequents of the normal form flow alone on the same surface of section as defined in figs.[fgresportrait]c , d . the same procedure is repeated in the second interval @xmath559 , where the motion is in general in the outward direction . the whole set of poincar consequents ( points @xmath270 gathered in this way are plotted in the 2d sections of the parallelepiped of fig.[fgarnolddif]c , along with the variations of the value of the normal form energy @xmath546 ( sampled more frequently ) which are shown in the third dimension . the details of the filling process of the various resonant chaotic layers located in the doubly - resonant domain are now clearly seen . in particular , we note that the chaotic orbit fills the whole separatrix layer of the initial resonance @xmath494 in a timescale much shorter than the one required for substantial drift in the @xmath41 direction . after a transient ` back and forth ' motion around @xmath577 , the orbit then moves slowly towards the value @xmath524 , where all important resonances overlap . in the intermediate time interval ( red ) , we clearly see the filling of the stochastic layers of both resonances @xmath495 and @xmath491 , while global transport is allowed by the normal form dynamics from one resonance to the other . as , however , the remainder effect causes a new motion of the orbit outwards ( i.e. towards higher values of @xmath41 ( green ) ) , the orbit is eventually captured at the resonance @xmath491 , and stays there until the end of the simulation at @xmath560 . it should be emphasized that the fact that the orbit moves in the outward direction at @xmath560 does not guarantee that there will be no further return inwards . in fact , we find that most orbits undergo several ` in - out ' cycles like the one described in fig.[fgarnolddif ] , before eventually abandoning the doubly - resonant domain . as an estimate , for @xmath528 we find that the number of cycles before a final exit from the doubly - resonant domain is of the order of 10 , while the total time required for this effect is of the order of @xmath578 to @xmath579 periods . furthermore , the probability of exit along one particular resonance decreases as the order of the resonance increases . this is expected , since the width of resonances scales with their order @xmath439 as @xmath580 , while the fast filling of the innermost chaotic domains where all the resonances overlap is nearly ergodic . finally , we point out that a visualization of the diffusion process like in fig.[fgarnolddif]c clearly suggests that the diffusion is driven by the intersections of the asymptotic manifolds of lower - dimensional objects ( like hyperbolic 2d tori ) all along the path in which the diffusion takes place . however , locating such tori , and studying their manifolds is a task that can not be accomplished by the use of the birkhoff normal form as above . on the other hand , the latter provides good initial conditions for a numerical search of such tori . this subject is proposed for future study . our final goal is to obtain numerical estimates of the value of the diffusion coefficient @xmath1 as well as its relation to the size @xmath2 of the optimal normal form remainder as @xmath0 is varied in the interval @xmath581 . to this end we implement the following numerical procedure : for any fixed value of @xmath0 , using the information from the fli maps , we first select 100 initial conditions corresponding to on a circle defined as in eq.([eres12 ] ) , where the radius is chosen equal to @xmath582 for such a choice of @xmath43 , the corresponding circle lies inside the resonance overlap domain , ensuring that the short time dynamics is dominated by the doubly - resonant normal form . however , in longer times all these orbits exhibit weakly chaotic diffusion . the complete set of initial conditions for one orbit on the circle @xmath583 are found by solving simultaneously for @xmath584 and @xmath585 the equation of the circle ( eq.([eres ] ) ) as well as an equation for the initial angle @xmath586 $ ] , where , for each initial condition , @xmath587 is chosen by visual inspection so as to correspond to an initial condition in the domain of each one of the main overlapping resonances . with @xmath430 , for ( a ) @xmath454 , ( b ) ) @xmath588 and ( c ) @xmath434 . the black points show the orbits consequents on the surface of section up to a time @xmath589 . all three orbits are diffusing outwards . the circle with radius @xmath590 is shown in pink . ] we then follow numerically these orbits for a time long enough so that the mean change of their radial distance from the center is large enough to allow for a reliable computation of the diffusion coefficient . let @xmath591 be the instantaneous value of the distance from the center for any such orbit . the quantity @xmath592 ^ 2 $ ] changes as an orbit slowly drifts from one circle to another . figure [ fgdifinout ] shows this effect for three orbits corresponding to the same initial angle @xmath587 but for three different values of @xmath0 , namely @xmath454 ( fig.[fgdifinout]a ) , @xmath588 ( fig.[fgdifinout]b ) , and @xmath434 ( fig.[fgdifinout]c ) . the orbits are shown by the black points on the section @xmath593 , @xmath480 , superposed as usually to the colored background of the fli map . the pink circles in each panel are the circles @xmath594 , where the orbits initial conditions lie . apart from an overall change of the size of the circle of initial conditions with @xmath0 , a simple visual comparison of the three panels suffices to conclude that they imply quite different diffusion rates of their depicted orbits . in all three panels , the orbits ( black points ) are shown up to a time @xmath589 , which is quite long compared to the time needed to fill the chaotic domain along the circle @xmath457 . however , when @xmath454 ( fig.[fgdifinout]a ) , the orbit s plot shows that the orbit exhibits no discernible transverse motion with respect to this circle , despite the fact that the orbit lies entirely within a rather strong chaotic domain ( yellow in the fli scale ) . on the other hand , when @xmath0 is raised to @xmath588 ( fig.[fgdifinout]b ) , the orbit is observed to create a small ring around its initial circle , implying that the diffusion is visible in this timescale . increasing @xmath0 still further ( @xmath434 , fig.[fgdifinout]c ) , causes now a rather fast diffusion , which leads to the orbit following clearly a preferential ` exit resonance ' , where the diffusion continues essentially as in the simple resonance case ( subsection 2.3.2 ) . and @xmath223 ( see text ) , using the new transformed canonical variables ( first and second panel ) , and @xmath595 and @xmath223 using the original canonical variables ( third and fourth panel ) , for two chaotic orbits of our chosen ensemble for @xmath528 ( upper and lower row).,title="fig : " ] and @xmath223 ( see text ) , using the new transformed canonical variables ( first and second panel ) , and @xmath595 and @xmath223 using the original canonical variables ( third and fourth panel ) , for two chaotic orbits of our chosen ensemble for @xmath528 ( upper and lower row).,title="fig : " ] a key remark , now , is the following : similarly to the case of the orbit of fig.[fgarnolddif ] , whose dynamical features were possible to unravel using the _ new , i.e. , transformed _ canonical variables after an optimal normalizing transformation , exploiting the same variables , instead of the original ones , allows to observe the random walk - like drift of one orbit in the action space _ in a much shorter integration time than by the use of the original variables . _ an example is given in fig.[fgdrift ] , for @xmath528 . we compute , via the optimal normalizing canonical transformation , a time sequence of the values of all the transformed canonical variables @xmath596 from the available sequences of values of the original variables @xmath597 along the numerical orbits . the four panels in each row show the time evolution , for one chaotic orbit on the circle @xmath430 , of the quantities i ) @xmath41 computed in the transformed canonical variables , ii ) @xmath598 computed in the transformed variables , iii ) @xmath595 computed in the original canonical variables , and iv ) @xmath223 computed in the original variables . we note immediately the gain by passing the data through the optimal normalizing transformation , namely the fact that this transformation absorbs all ` deformation ' effects , allowing to see the very slow drift due to the weakly chaotic diffusion in a timescale @xmath599 . in fact , the quantity @xmath41 can only be computed in the transformed canonical variables , in which , for both orbits , it undergoes variations of the order @xmath600 . in comparison , the analog of @xmath41 in the original variables , i.e. , @xmath595 , undergoes variations in the second digit , and the corresponding time evolution is dominated by @xmath28 oscillations , which completely hide the slow drift process in the radial direction with respect to the central doubly resonant point . the comparison is even more straightforward in the variables @xmath223 computed by the transformed and by the original action variables . in the former , we can clearly see the drift phenomenon for both orbits , which results in a slow change of the value of @xmath223 ( which is an approximate integral ) at the fifth digit . in contrast , this phenomenon is completely hidden when @xmath223 is computed in the original variables , since the corresponding plot is dominated by oscillations of at least one order of magnitude larger amplitude than the drift effect . ( see text ) , in our ensemble of numerical data for @xmath528 , when computed by use of the new transformed canonical variables ( left ) , or the original canonical variables ( right ) . the evolution is shown up to the time @xmath601 . ] in order , now , to measure the value of the diffusion coefficient , using the data from all 100 orbits , we define the mean square deviation : @xmath602 where @xmath603 , and @xmath604 stands for any of the four quantities shown in fig.[fgdrift ] . plotting @xmath605 against the time @xmath16 allows to estimate the diffusion coefficient . figure [ fgdif008 ] shows an example of this calculation , setting @xmath606 equal to @xmath223 in the transformed variables ( left panel ) , or the original variables ( right panel ) . we note again that it becomes possible to observe the diffusion in a timescale @xmath607 using the ensemble of data in the transformed variables , while this time is quite short to reveal any linear trend of @xmath608 with the time @xmath16 in the original variables . in fact , in the original variables it was possible to measure reliably the diffusion coefficient only after an integration time @xmath609 . furthermore , this time increases even more for smaller values of @xmath0 . ( upper set of points ) and @xmath610 ( lower set of points ) on the optimal normal form remainder @xmath611 , using numerical data from the integration of orbits ( see text ) . the points correspond to the values of @xmath0 ( from left to right ) 0.003 , 0.004 , 0.005 , 0.007,0.008 , 0.01 , 0.012 , 0.013 , 0.015 , 0.018 , 0.020 . the straight lines represent the power - law fits @xmath612 ( upper ) and @xmath613 ( lower ) . ] figure [ fgdifcfrem ] shows the final result . computing , as indicated above , the diffusion coefficients @xmath614 and @xmath610 in the transformed canonical variables , for eleven different values of @xmath0 as noted in the caption , we also use the data from fig.[fgremopt ] , whereby we obtain the optimal remainder value @xmath2 for the same values of @xmath0 ( from the minima of the curves of fig.[fgremopt ] ) . we then plot @xmath614 and @xmath610 against @xmath2 in a log - log scale . despite some scatter , the correlation of both independent estimates of the diffusion coefficient with @xmath611 can be described as a power - law . the power - law exponents found by best - fitting are @xmath615 for the data of @xmath614 and @xmath616 for the data of @xmath610 . in these best fittings we excluded the two points for @xmath455 and @xmath454 , since the value of the optimal remainder found by extrapolation is uncertain for these values of @xmath0 . however , we note that the corresponding points in fig.[fgdifcfrem ] are still very close to the fitting law found by the remaining data . the exponents found in fig.[fgremopt ] are not far from the theoretical estimate @xmath617 derived in section 2 ( eq.([difcfest ] ) ) . however , we have made various trials to determine @xmath426 via alternative definitions of the diffusion coefficient , and we always find estimations of @xmath426 somewhat larger than 2 . we thus conjecture that this difference from @xmath617 is a real effect ( not due to numerical uncertainties ) , which , however , requires a more detailed theory to interpret . on the other hand , the corresponding analysis for simple resonances ( subsection 2.3.2 ) as well as the numerical results of @xcite indicate that the steepening of the power law in simple resonances of order not smaller than @xmath121 is quite substantial , leading closer to @xmath33 . in the latter case , another independent example @xcite yields @xmath618 . the issue of how exactly to quantify the steepening of the power - law remains open . we examined in detail the phenomenon of weak chaotic diffusion in doubly or simply resonant domains of hamiltonian systems of three degrees of freedom satisfying the necessary conditions for the holding of the nekhoroshev theorem . the aim was to determine a quantitative relation between the diffusion coefficient @xmath1 and the size of the optimal remainder @xmath2 of a resonant normal form constructed according to the requirements of the analytical part of the nekhoroshev theorem . our main results are the following : \1 ) we propose an efficient algorithm for hamiltonian normalization , which is implemented as a computer algebraic program performing expansions up to a high order . we explain the practical aspects of this algorithm , and show how it can be used in order to compute i ) the optimal normalization order @xmath158 as a function of the small parameter @xmath0 , and ii ) an estimate of the size of the remainder @xmath2 at the order @xmath158 . the dependence of @xmath158 on @xmath0 is found to be an inverse power - law with an exponent in agreement with theory . \2 ) we construct estimates on the speed of diffusion in doubly resonant domains . to this end , we examine first the dynamics under the hamiltonian flow induced by the normal form alone ( i.e. neglecting the remainder ) . the role of the convexity conditions assumed for the original hamiltonian is analyzed in the context of the normal form dynamics . we then discuss the influence of the remainder on dynamics . estimates on the value of the diffusion coefficient @xmath1 are quantified by considering a ` random walk ' model for the slow drift of the value of the normal form energy due to the remainder . the final prediction is a power - law estimate @xmath619 with @xmath389 in doubly resonant domains . \3 ) we perform detailed numerical experiments aiming to test the above predictions , employing the same hamiltonian model as in @xcite as well as the ` fli map ' method . using the information from the computed normalizing canonical transformations , we propose a convenient set of variables in which the arnold diffusion in the doubly resonant domains is clearly visualized . furthermore , using ensembles of chaotic orbits , we make two independent numerical calculations of the diffusion coefficient @xmath1 for various values of @xmath0 . the relation between @xmath1 and @xmath2 found by the two calculations is @xmath620 and @xmath621 respectively . \4 ) finally , we make some theoretical estimates on the relation between @xmath1 and @xmath2 in simply resonant domains . in this case , we combine the basic theory developed in @xcite together with estimates given in @xcite regarding the dependence of the size of the separatrix splitting on the optimal normal form remainder in simply resonant domains . we are thus led to the prediction @xmath622 , where @xmath357 , or @xmath359 , holding for all simple resonances of order higher than @xmath121 , where @xmath121 is defined in eq.([kprime ] ) . the latter result interprets the results obtained in an earlier study @xcite by purely numerical means . + + acknowledgements : * we thank two anonymous referees for a thorough revision of our manuscript , with many constructive suggestions , as well as prof . g. contopoulos for careful reading of the manuscript . c.e . acknowledges fruitful discussions with the group of c. froeschl , m. guzzo and e. lega . arnold , v.i . , 1963 : _ russ . math . surveys _ * 18 * , 9 . arnold , v.i . , 1964 : _ sov . dokl . _ * 6 * , 581 . benettin , g. , galgani , l. , and giorgilli , a. : 1985 , _ cel . _ * 37 * , 1 . benettin , g. , and gallavotti , g. : 1986 , * j. stat . phys . * * 44 * , 293 . benettin g. , fass f. , guzzo m. : 1998 , _ regular chaot . _ , * 3 * , 56 . benettin , g. : 1999 , in a. giorgilli ( ed . ) ` hamiltonian dynamics . theory and applications ' . notes math . _ * 1861 * , 1 . cachucho , f. , cincotta , p.m. , and ferraz - mello , s. : 2010 , _ cel . astron . _ * 108 * , 35 . celletti , a. , and ferrara , l. : 1996,_cel . astron . _ * 64 * , 261 . chirikov , b.v . : 1979 , _ phys . * 52 * , 263 . chirikov , b.v . , and vecheslavov , v.v . : 1985 , _ physica _ * 71 * , 243 . cincotta , p. : 2002 , _ new astronomy reviews _ * 46 * , 13 . cincotta , p. , giordano , c. , mestre , m. , and efthymiopoulos , c : 2013 , in preparation . contopoulos , g. : 1966 , _ bull . cnrs _ * 2 * , 223 . contopoulos , g. : 2002 , _ order and chaos in dynamical astronomy _ , springer , berlin . dumas , h.s . , and laskar , j. : 1993 , _ phys . lett . _ * 70 * , 2975 . efthymiopoulos , c. , contopoulos , g. and voglis , n. : 1998 , in benest , d. , and froeschl , c. ( eds ) _ discrete dynamical systems _ , gordon and breach science publishers , pp.91 - 106 . efthymiopoulos , c. , giorgilli , a. , and contopoulos , g : 2004 , _ j. phys . a math . * 37 * , 10831 . efthymiopoulos , c. : 2005 , _ cel . astron . _ * 92 * , 29 . efthymiopoulos , c. , and sandor , z. : 2005 , _ mon . not . r. astron . soc . _ * 364 * , 253 . efthymiopoulos , c. : 2008 , _ cel . astron . _ * 102 * , 49 . ferraz - mello , s. : 2007 , _ canonical perturbation theories . degenerate systems and resonance_. springer ( new york ) . froeschl , c. , guzzo , m. , and lega , e. : 2000 , _ science _ * 289 * ( 5487 ) , 2108 . froeschl , c. , guzzo , m. , and lega , e. : 2005 , _ cel . astron . _ * 92 * , 243 . gelfreich , v. , sim , c. , and vieiro , a. : 2013 , _ physica d _ * 243 * , 92 . giordano , c.m . , and cincotta , p.m. : 2004 , _ astron . astrophys . _ * 423 * , 745 . giorgilli a. , delshams a. , fontich e. , galgani l. , sim c. : 1989 , _ j. differ . _ , * 77 * , 167 . giorgilli a. , and locatelli , u. : 1997 , _ zamp _ , * 48 * , 220 . giorgilli a. , and skokos , c. : 1997 , _ astron . _ , * 317 * , 254 . giorgilli , a. : 1999 , in : c. simo ( ed . ) , _ hamiltonian systems with three or more degrees of freedom _ , kluwer , dordrecht . giorgilli , a : 2002 , _ notes on exponential stability of hamiltonian systems _ , in dynamical systems . part i : hamiltonian systems and celestial mechanics , pubblicazioni della classe di scienze , scuola normale superiore , pisa . giorgilli , a. , locatelli , u. , and sansoterra , m. : 2009 , _ cel . * 104 * , 159 . guzzo , m. , lega , e. , and froeschl , c. : 2002 , _ physica d _ * 163 * , 1 . guzzo , m. , lega , e. , and froeschl , c. : 2005 , _ dis . b _ * 5 * , 687 . guzzo , m. , lega , e. , and froeschl , c. : 2006 , _ nonlinearity _ * 19 * , 1049 . guzzo , m. , lega , e. , and froeschl , c. : 2011 , _ chaos _ * 21 * , 033101 . kaneko , k. , and konishi , t. : 1989 , _ phys . rev . a _ * 40 * , 6130 . laskar , j. : 1993 , _ physica _ * d67 * , 257 . lega , e. , guzzo , m. , and froeschl , c. : 2003 , _ physica d _ * 182 * , 179 . lega , e. , froeschl , c. , and guzzo , m. : 2007 , _ lect . notes phys . _ * 729 * , 29 . lega , e. , guzzo , m. , and froeschl , c. : 2009 , _ cel . astron . _ * 104 * , 191 . lega , e. , guzzo , m. , and froeschl , c. : 2010a , _ cel . * 107 * , 129 . lega , e. , guzzo , m. , and froeschl , c. : 2010b , _ cel . astron . _ * 107 * , 115 . lhotka , ch . , efthymiopoulos , c. , and dvorak , r. : 2008 , _ mon . not . r. astron . soc . _ * 384 * , 1165 . liechtenberg , a.j . , and lieberman , m.a . : 1992 , _ regular and chaotic dynamics _ , springer , berlin . lochak , p. : 1992 surv . _ * 47 * , 57 . mather , j. : 2004 , _ j. math . _ * 124 * , 5275 . morbidelli , a. : 2002 , _ modern celestial mechanics . aspects of solar system dynamics _ , taylor and francis , london . morbidelli , a. , and guzzo , m. : 1997 , _ cel . astron . _ * 65 * , 107 . morbidelli , a. , and giorgilli , a. : 1997 , _ physica d _ * 102 * , 195 . neishtadt , a.i . : 1984 , _ mech . _ * 48 * , 133 . nekhoroshev , n.n . : 1977 , _ russ . surv . _ * 32(6 ) , 1 . pavlovic , r. , and guzzo , m. : 2008 , _ mon . not . r. astron . soc . _ 384 * , 1575 . poincar , h : 1892 , _ mthodes nouvelles de la mcanique cleste _ , gautier - vilard , paris . pshel , j. : 1993 , _ math . z. _ * 213 * , 187 . rosenbluth , m. , sagdeev , r. , taylor , j. , and zaslavskii , m. : 1966 , _ nucl . fusion _ * 6 * , 217 . sim , c. , and valls , c. : 2001 , * nonlinearity * * 14 * , 1707 . tennyson , j. : 1982 , _ physica d _ * 5 * , 123 . skokos , c. , contopoulos , g. , and polymilis , c. : 1997 , _ cel . astr . _ * 65 * , 223 . wood , b.p . , lichtenberg , a.j . , and lieberman , m.a . : 1990,_phys . a _ * 42 * , 5885 . the quadratic form @xmath623 given by eq.([z02 ] ) can be written as : @xmath624 where @xmath625 is a @xmath626 matrix whose first and second line are given by @xmath627 and @xmath628 respectively . since the matrix @xmath66 is real symmetric , it can be writen in the form @xmath629 , where @xmath630 , with @xmath631 the eigenvalues of @xmath66 , while @xmath632 is an orthogonal matrix with columns equal to the normalized eigenvectors of @xmath66 . using the above expression for @xmath66 , eq.([z02matr ] ) resumes the form @xmath633 where @xmath634 is a @xmath626 matrix . writing @xmath623 as @xmath635 , and denoting by @xmath636 the elements of @xmath606 , the discriminant @xmath637 is given by : @xmath638\ ] ] since we have assumed ( subsection 2.1 ) that either all three eigenvalues @xmath639 have the same sign , or two of them have the same sign and one is zero , by eq.([z02anal ] ) we have that @xmath640 . that is , the quadratic form @xmath623 is positive definite .	a detailed numerical study is presented of the slow diffusion ( arnold diffusion ) taking place around resonance crossings in nearly integrable hamiltonian systems of three degrees of freedom in the so - called ` nekhoroshev regime ' . the aim is to construct estimates regarding the speed of diffusion based on the numerical values of a truncated form of the so - called remainder of a normalized hamiltonian function , and to compare them with the outcomes of direct numerical experiments using ensembles of orbits . in this comparison we examine , one by one , the main steps of the so - called analytic and geometric parts of the nekhoroshev theorem . thus : i ) we review and implement an algorithm @xcite for hamiltonian normalization in multiply resonant domains which is implemented as a computer program making calculations up to a high normalization order . ii ) we compute the dependence of the optimal normalization order on the small parameter @xmath0 in a specific model and compare the result with theoretical estimates on this dependence . iii ) we examine in detail the consequences of assuming simple convexity conditions for the unperturbed hamiltonian on the geometry of the resonances and on the phase space structure around resonance crossings . iv ) we discuss the dynamical mechanisms by which the remainder of the optimal hamiltonian normal form drives the diffusion process . through these steps , we are led to two main results : i ) we construct in our concrete example a convenient set of variables , proposed first by benettin and gallavotti @xcite , in which the phenomenon of arnold diffusion in doubly resonant domains can be clearly visualized . ii ) we determine , by numerical fitting of our data the dependence of the local diffusion coefficient @xmath1 on the size @xmath2 of the optimal remainder function , and we compare this with a heuristic argument based on the assumption of normal diffusion . we find a power law @xmath3 , where the constant @xmath4 has a small positive value depending also on the multiplicity of the resonance considered . hamiltonian systems ; arnold diffusion ; normal forms ; nekhoroshev theorem .
we use the same notation as introduced in the main text . the floquet propagator of a on - resonance double - kicked rotor is given by @xmath103 with eigenstates and eigenvalues denoted by @xmath104 here @xmath52 represent the bloch eigenstates of @xmath105 with bloch phase @xmath106 , due to momentum - space periodicity of period @xmath107 for @xmath108 . using the bloch theorem , one rewrites @xmath109 as @xmath110 here @xmath111 holds for all momentum eigenstates @xmath112 . we adopt the normalization convention @xmath113 . with this normalization convention , it can be shown that the identity operator in terms of the bloch eigenstates is given by @xmath114 and their inner products are given by @xmath115 which implies that the initial states we considered in the main text [ see also eq . ( [ specialstate ] ) below ] are normalized to unity . note that we have chosen a basis of bloch eigenstates , @xmath116 , which are set to be periodic functions of @xmath38 , meaning that @xmath117 for all @xmath6 . also , the overall phase convention of @xmath116 can be chosen such that it is a locally single - valued function of @xmath38 and @xmath6 . next , we consider the states involved in the adiabatic cycle described in the main text . such states are of the form @xmath118 where @xmath119 takes values from @xmath47 to @xmath120 as explained in the main text and @xmath121 . we begin our derivation of quantized transport in momentum space by taking momentum expectation value of such states , @xmath122 we insert two copies of eq . ( [ identity ] ) , one in the middle of each of the inner products of the right hand side . then , making use of eqs . ( [ specialstate ] ) , ( [ innerprod ] ) and ( [ blochform ] ) , we arrive at @xmath123 we then rewrite @xmath12 as a derivative in @xmath38 acting on the exponential term and perform an integration by parts to shift this derivative off the exponential and onto the other @xmath38-dependent terms . further , we split @xmath12 into @xmath124 , where @xmath125 takes integer values @xmath126 , and @xmath119 takes all integer values from @xmath127 to @xmath128 . then , after making use of the fact that @xmath129 for all momenta @xmath130 and all @xmath131 , one invokes poisson s summation formula @xmath132 and finds @xmath133 where @xmath134 and we have made use of the fact that @xmath135 . summarizing , we have shown that @xmath136 note that @xmath137 are the eigenstates of the reduced floquet matrix @xmath138 with @xmath139 and @xmath140 . next , we consider the difference in momentum expectation value between consecutive states in the adiabatic cycle , namely , @xmath141 for both terms in the above equation , the expectation values are given by eq . ( [ p_expectatn ] ) with @xmath73 and @xmath142 respectively . next one may express @xmath143 in terms of @xmath144 to the first order in @xmath145 , i.e. , @xmath146 where @xmath147 this perturbation theory can be performed because ( i ) our bloch eigenstates can be chosen as continuous in @xmath6 for any fixed @xmath38 and ( ii ) in the adiabatic limit @xmath148 , @xmath149 and hence only first - order matters . it is now seen that @xmath150 is expressed as the eigenstate ( correct to the first order in @xmath151 ) of the reduced floquet matrix with @xmath152 , plus a correction . note also that the expression @xmath153 in the correction term can be explicitly calculated by @xmath154 via @xmath155 finally , we use the feynman - hellmann - like formula @xmath156 and its adjoint . substituting eq . ( [ feynhell_like ] ) , its adjoint , and eq . ( [ 1st_order_state ] ) into eq . ( [ small_p_change ] ) and keeping only first order terms in @xmath151 , we obtain @xmath157 where @xmath84 is identified as the berry curvature @xmath158 , @xmath86 and @xmath87 on @xmath6 and @xmath38 all suppressed for brevity . the total change in the expectation value of @xmath159 from @xmath88 to @xmath80 is denoted by @xmath160 . because @xmath161 , one finds the total momentum change over the entire adiabatic cycle , i.e. , @xmath162 where we have used @xmath25 and @xmath92 is exactly the chern number of the @xmath41th floquet band defined earlier , whose equivalent expression is @xmath163 this completes our derivation of quantization of transport in momentum space . here we present some details regarding how the chern numbers of fig . 2 of the main text are calculated . in terms of a line integral , the chern number is given by @xmath164 where @xmath165 , @xmath166 ( normalized on @xmath27 momentum components ) . we first mention that this expression is equivalent to @xmath167 , where @xmath168 , and @xmath169 , with @xmath170 . to see this , one needs only to substitute @xmath169 into the chern number expression above and write the integral as the sum of 4 terms , each on one edge of the @xmath39 brillouin zone . the two terms containing @xmath171 immediately reduce to the form @xmath172 because @xmath173 is not a function of @xmath6 . the other two terms containing @xmath174 will also reduce to the similar form @xmath175 after one application of the product rule of differentiation and noting that two of the resulting four terms cancel . this then leaves us with @xmath176 . we can now work on @xmath177 in terms of @xmath178 to make the link with the main text more concrete , though it should be clear that it is entirely valid to work with @xmath179 instead . in the remainder of this section , we will derive the surface integral formula for the chern number as it was used in the main text and then mention some details on our numerical calculation . the surface integral for @xmath177 which involves the berry curvature as shown in the main text is obtained from the line integral via the steps shown here . first , a single application of stoke s theorem leaves us with @xmath180 \nonumber \\ & = & \frac{i}{2\pi } \int _ { 0 } ^{2\pi } \int _ { 0 } ^{2\pi } d\phi d\alpha \left [ \left ( \frac{\partial}{\partial \phi}{\left\langle \bar{u}_{n}(\phi,\alpha)\right\| } \right ) \frac{\partial}{\partial \alpha}{\left\| \bar{u}_{n}(\phi,\alpha)\right\rangle } -\left(\frac{\partial}{\partial \alpha}{\left\langle \bar{u}_{n}(\phi,\alpha)\right\|}\right ) \frac{\partial}{\partial \phi}{\left\| \bar{u}_{n}(\phi,\alpha)\right\rangle}\right ] . \nonumber \\ \label{tedious_formula}\end{aligned}\ ] ] we then insert one copy of the @xmath68 identity matrix of the form @xmath181 between each of the two inner products , where we have refrained from writing the functional dependence of @xmath178 on @xmath39 for brevity . the term with @xmath182 in one inner product will cancel that in the other inner product because @xmath183 and similarly for @xmath184 . this leads to @xmath185.\ ] ] to proceed , we note that the states @xmath186 are eigenstates of the reduced floquet matrix @xmath187 with @xmath139 , and where @xmath37 is the `` on - resonance double kicked rotor model '' floquet operator defined in the main text . from the eigenvalue equation @xmath188 , we may take a derivative with respect to @xmath38 on both sides , take the inner product with @xmath189 from the left on both sides , and rearrange to obtain the feynman - hellmann - like formula @xmath190 which has adjoint given by @xmath191 similar expressions also hold for the @xmath171 derivative . we substitute the feynman - hellman - like formulas and their adjoints for both @xmath38 and @xmath6 into eq . ( [ tedious_formula ] ) and obtain @xmath192 where latexmath:[\[b _ { n } ( \phi,\alpha ) = i \sum _ { n'=1,\neq n } ^{n } \bigg\ { \frac{\langle \bar{u}_{n}\|\frac{\partial \bar{u}^{\dagger}}{\partial \phi}\|\bar{u}_{n'}\rangle\langle \bar{u}_{n ' } is the berry curvature . this completes the derivation of the surface integral for @xmath177 featured in the main text . we now mention some brief details on how we numerically calculate the chern numbers . we chose eq . ( [ chern_surface_integral ] ) to calculate the chern numbers rather than eq . ( [ chern_num_defn ] ) because the former method allows us to use any phase convention for the @xmath178 eigenstates ( a quick check shows that the berry curvature is invariant with respect to replacing each eigenstate with itself multiplied by a global phase ) . first , we discretize the @xmath39 brillouin zone into a sufficiently fine grid . we then numerically work out the integrand of eq . ( [ chern_surface_integral ] ) at each point on the grid and numerically integrate to obtain the chern number . the different parts of the integrand are obtained as follows . we numerically evaluate the @xmath68 elements of the @xmath194 matrix via eq . ( [ reduced_floquet ] ) . the infinite sum on the right hand side there does not present a problem because @xmath195 decays rapidly as @xmath196 increases , so we can truncate the sum at some suitable point with practically no loss of accuracy . each term @xmath195 is efficiently calculated via fast fourier transform ( fft ) . we then numerically diagonalize and obtain the @xmath27-element vectors representing the @xmath178 eigenstates as well as the eigenphases @xmath197 . lastly , we obtain the elements of the @xmath198 matrices @xmath199 and @xmath200 . this is done by acting @xmath174 and @xmath171 on both sides of eq . ( [ reduced_floquet ] ) , simplifying the expression on the right hand side by hand , and then obtaining the numerical values via fft . the infinite sums here may again be truncated at a suitable level with no appreciable loss in accuracy . to motivate potential experimental interest , in this section we consider the effects of of nonzero quasimomentum variable @xmath15 , small perturbations to effective planck constant @xmath31 , and inevitable imperfections in actual experimental implementations on the quantization of transport in momentum space . to consider the effects of nonzero @xmath15 , we begin from the double - kicked rotor model floquet propagator , @xmath201 which describes a kicked particle moving on a line . as explained in @xcite , the periodicity of the floquet propagator in @xmath9 allows us to map the dynamics of a kicked particle on a line onto the dynamics of a fictitious ensemble of kicked particles each on a circle , with every such particle representing the dynamics for a different @xmath15-component of the kicked particle on the line . each such particle on the circle is thus referred to as a @xmath15-rotor . the @xmath15 rotors may then be evolved separately from one another and recombined appropriately at the end to recover the time - evolved state of the actual particle on the line . the different @xmath15-rotors evolve independently of one another under the action of a @xmath15 floquet propagator @xmath202 where @xmath203 is the @xmath15-rotor s momentum operator with eigenstates @xmath204 and momentum eigenvalues @xmath205 with @xmath206 . now , imposing the main quantum resonance condition @xmath207 , and writing @xmath18 , @xmath208 and @xmath209 , the @xmath15 floquet propagator becomes @xmath210}e^{-i({k _ { e } } / { \hbar _ { e}})\mathrm{cos}({q})}e^{-i[(\hat{n}_{e}+\beta{\hbar _ { e } } ) ^{2 } / 2 { \hbar _ { e } } ] } e^{-i ( { k _ { e } } / { \hbar _ { e}})\mathrm{cos}({q}+\alpha)}. \label{betapropagator}\end{aligned}\ ] ] the rescaled momentum eigenvalues are then given by @xmath211 , @xmath206 . if we set @xmath212 , the momentum operator has eigenvalues of only integer multiples of @xmath31 and we recover the on - resonance double - kicked - rotor - model floquet propagator described in eq . ( 1 ) of the main text . for nonzero @xmath15 , @xmath213 in eq . ( [ betapropagator ] ) lacks translational invariance in momentum , so the entire derivation in the previous section will not apply and we hence do not expect to see quantized momentum transport in the @xmath15-rotor dynamics . however , for small @xmath15 values , it is reasonable to expect that evolution of the special initial state in eq . ( [ specialstate ] ) will still show quantized transport up to a good approximation . we find that this is indeed the case , as evidenced in fig . 5(a ) , where we see that for @xmath15 values up to @xmath214 , the quantized transport still survives for @xmath215 , @xmath94 . experimentally , however , the initial states prepared are typically a mixture of states with @xmath15 values following a narrow gaussian distribution peaked at @xmath14 . hence , to gauge the experimental feasibility of our proposal , we computed the transport values for a range of @xmath15 values near @xmath47 and took the average of these values weighted with a gaussian peaked at @xmath47 . we found that for a gaussian distribution with standard deviation of @xmath214 , the change in momentum expectation value divided by @xmath216 for the 3 bands at @xmath217 for a 100-period adiabatic cycle are @xmath218 , @xmath219 and @xmath218 for bands 1,2 and 3 , respectively . in other words , the quantization of transport is still observable even with a realistic spread over @xmath15 values . this indicates that the small @xmath15 spread in current cold - atom experiments based on bose - einstein condensates should not hinder the observation of quantized transport in momentum space . next , we briefly consider the effects of small deviations in the planck constant @xmath31 for the same 3-band case considered above with @xmath220 and @xmath221 . certainly , even a slight change in @xmath31 should cause the floquet - band structure to change entirely ( recall that the floquet band structure depends on whether @xmath31 is rational or irrational ) , but physical observables should not be as sensitive as the band structure itself . indeed , as seen in fig . 5(b ) , the quantized transport is still observable for deviations in @xmath31 up to @xmath222 . @xmath223{fig1a - supp.pdf } & \includegraphics[width=80 mm , height=!]{fig1b - supp.pdf}\\ \end{array}$ ] @xmath223{fig2a - supp.pdf } & \includegraphics[width=80 mm , height=!]{fig2b - supp.pdf}\\ \end{array}$ ] finally , we consider the effects of practical imperfections in experiments . to take into account the difficulty of superposing a large number of different momentum eigenstates in actual experiments , we consider a perturbed version of the initial state @xmath224 by setting to zero all probability amplitudes outside the range @xmath225 $ ] and multiplying the resulting state by an appropriate constant for normalization . this does not alter the original state significantly because its probability is tightly concentrated about @xmath226 ( more than @xmath227 % of the original state s probability lies within @xmath225 $ ] ) . with this truncated state as our starting point , we consider two noise models . in the first model , we introduce an uncertainty in the relative phases of the 11 surviving momentum states in the superposition by multiplying each one by exp@xmath228 $ ] where @xmath229 is a random variable uniformly distributed in [ 0,1 ) . we then assume that the probability amplitude of each constituent momentum state is scaled by the term [ @xmath230 and then multiply all states by an appropriate constant for normalization . in our second noise model , we assume that in the @xmath119-th period , the value of the phase shift @xmath6 is given by @xmath2312\pi/100 $ ] . we now present in fig . 6 two simple numerical examples for band 1 and band 3 , with @xmath95 , @xmath232 , and @xmath233 . it is seen that in both models , the total momentum transport over one cycle ( divided by @xmath216 ) , when averaged over the 1000 realizations of the random noise , changes only slightly from the quantized value ( @xmath234 ) despite the rather strong noise . the fair robustness to imperfection to initial state shown in fig . 6(a ) is partially due to the fact that our initial state is dominated by very few momentum components . indeed , the fluctuations in the total momentum transport from realization to realization , as manifested by the plotted error bars for band 3 , are much smaller than those for band 1 , consistent with the fact that our initial state for band 3 turns out to be more localized than that for band 1 . results in fig . 6(b ) illustrate that slow fluctuations in @xmath6 ( noise amplitude rather comparable to @xmath235 ) during the adiabatic process do not severely affect the quantization . as seen from fig . 6(b ) , the total momentum transport averaged over 1000 noise histories is only slightly shifted from the quantized value ( @xmath234 ) , and fluctuations represented by the error bars are also rather small . this confirms that the fine details of an adiabatic process are largely irrelevant . increases , for @xmath236 . the horizontal lines represent the bands and the numbers written on them are their chern numbers . vertical lines connecting two horizontal lines represent degeneracies occurring between the two associated bands . note the unusually big changes of chern numbers across some critical points.,width=529,height=272 ] in this section , we wish to emphasize that our topological characterization of the floquet bands using chern numbers is equally applicable to cases with more bands . here we show a case with 7 bands , with @xmath237 . we present in fig . 7 the chern number results as the kicking strength increases . the associated chern numbers are found to be symmetric about the central band , so we need only display 4 bands . the same qualitative features as we observed in the 3-band case are seen here . in particular , all bands other than the central band have chern number @xmath234 for low kicking strength , with the central band having a chern number equal to the negative of the sum of the chern numbers of all other bands . at band collisions , the chern number of each band jumps by a multiple of @xmath238 for @xmath239 . for this reason we see that the chern numbers can rapidly become very large integers , which is a quite interesting feature for our system . we have also carried out adiabatic transport studies for this case . due to the smaller band gaps for the 7-band case , we need to have longer @xmath6-cycles ( i.e. , @xmath6 must be varied from @xmath47 to @xmath24 over a larger number of periods ) in order for effective adiabatic following to take place . experimentally speaking , this case is hence of less interest . however , in our numerical studies , other than the longer adiabatic cycles , the results are qualitatively the same as the 3-band case presented in the main text . in this last section , we provide some further examples of distributions in momentum space after the adiabatic cycles proposed in the main text . we choose @xmath240 for @xmath241 , as we did in fig . 3 of the main text . figure 8 shows the initial and final distributions after a single adiabatic cycle for initial wannier states prepared on bands 1 and 2 . figure 9 shows the distribution in momentum space after completion of a second and third adiabatic cycle for an initial wannier state prepared on band 3 ( see fig . 3 of the main text ) .	though topological aspects of energy bands are known to play a key role in quantum transport in solid - state systems , the implications of floquet band topology for transport in momentum space ( i.e. , acceleration ) are not explored so far . using a ratchet accelerator model inspired by existing cold - atom experiments , here we characterize a class of extended floquet bands of one - dimensional driven quantum systems by chern numbers , reveal topological phase transitions therein , and theoretically predict the quantization of adiabatic transport in momentum space . numerical results confirm our theory and indicate the feasibility of experimental studies . in both classical mechanics and quantum mechanics , position and momentum variables form a conjugate pair and can hence be treated on the same footing from a phase space perspective . the real physical world , however , does not have position - momentum symmetry . for example , energy bands of a solid are formed because electronic hamiltonians are periodic in position but not in momentum . due to such unequal roles of position and momentum , the mapping of quantum transport phenomena from position space to momentum space is typically nontrivial but , where possible , may lead to important insights and unforeseen opportunities . anderson localization , for instance , was first discovered as a seminal result of quantum transport in position space . its analog in momentum space was later found to be behind the intriguing phenomenon of dynamical localization " @xcite . this mapping has stimulated fruitful studies of anderson transition in driven cold - atom systems @xcite . as a second example , ratchet transport , namely , directed transport under a zero mean force in position space , has been mapped to momentum space as well , leading to the finding of ratchet accelerators ( ra ) @xcite . energy - band topology is of fundamental interest to studies of quantum transport in condensed - matter physics . the issue to be addressed here is whether transport in momentum space ( i.e. , acceleration ) can be connected with band structures due to momentum space periodicity . at first glance this sounds impossible because , with the ( non - relativistic ) kinetic energy being a quadratic function of momentum , a realistic hamiltonian is never a periodic function of momentum . however , for systems periodically driven by impulsive fields , the floquet operator can still be periodic in momentum , thus forming floquet ( quasi - energy ) bands . the topology of such floquet bands then provides a useful tool in characterizing topological phase transitions in driven quantum systems @xcite . how floquet - band topology is manifested in acceleration then becomes an intriguing question . using a one - dimensional ra model inspired by existing cold - atom experiments , we show in this work that the floquet bands , defined on a 2-torus ( formed by one bloch phase and one experimentally tunable parameter ) may be characterized by chern numbers . we then theoretically show that adiabatic transport in _ momentum space _ can be quantized according to these topological numbers . unlike quantized adiabatic pumping in position space @xcite , there does not exist a general ( system - independent ) flux operator in momentum space and hence the found quantization is about a quantized net change of momentum expectation value , rather than a pumping of particles through a cross section . finally , though adiabatic manipulation of floquet states is often considered to be subtle @xcite , our numerical results confirm our theory and indicate that quantized momentum space transport can be observed in a wide regime , by simply scanning one system parameter in a relatively small number of discretized steps . breakdown of quantization in momentum space transport may then be considered as a diagnostic tool for detecting non - adiabatic effects in floquet state manipulation . to physically realize the main physics we need three ingredients : momentum space periodicity , an experimentally tunable periodic parameter and well - gapped floquet bands . to be specific we consider a ra model @xcite , which is based on an atom - optics realization of a double - kicked rotor system @xcite . the ra hamiltonian is given by @xmath0 , with a corresponding floquet propagator @xcite @xmath1 , where all quantities are properly scaled and hence in dimensionless units , @xmath2 and @xmath3 are canonical coordinate and momentum operators for cold atoms , and the @xmath4 kicks are of period @xmath5 , experimentally implemented by two optical - lattice potentials mutually phase - shifted by @xmath6 , with equal strength @xmath7 , and a time lag @xmath8 @xcite . note that later @xmath6 will be adiabatically tuned . because the potential function is periodic in @xmath9 , momentum eigenstates take eigenvalues @xmath10 , where @xmath11 $ ] is a conserved quasi - momentum variable and @xmath12 is an integer . to yield a floquet operator periodic in momentum despite the @xmath13 term in the hamiltonian , we set @xmath14 , which may be approximately implemented by considering a bose - einstein condensate whose coherence width spans across many optical lattice constants @xcite . effects of nonzero @xmath15 values will be discussed in appendix @xcite . if we now impose the quantum resonance condition @xmath16 that has been one experimental subject @xcite , we obtain an `` on - resonance double - kicked - rotor model '' @xcite , with the floquet propagator reducing to @xmath17 where @xmath18 is an effective planck constant , @xmath19 , and a rescaled momentum operator @xmath20 . from here on , momentum exclusively refers to @xmath21 and we denote momentum eigenstates by @xmath22 , which has eigenvalue @xmath23 and is periodic in @xmath9 with period @xmath24 . equation ( [ beta_rotor ] ) indicates that if @xmath25 , with @xmath26 and @xmath27 being integers , then the floquet operator @xmath28 is perfectly periodic in momentum space with a period of @xmath29 . according to bloch s theorem , this momentum space periodicity leads to floquet bands . indeed , in the case of @xmath30 , the corresponding floquet band structure @xcite resembles hofstadter s butterfly @xcite , with @xmath31 here identified as an analog of the magnetic flux . it is such a remarkable resemblance ( which also hints desirable band - gap features ) that motivated us to connect the topological aspects of the floquet bands with momentum - space transport . @xmath32 { fig1a.png } \hspace{-0.cm } \includegraphics[width=60 mm , height=52 mm ] { fig1b.png } \end{array}$ ] the floquet band structure may be characterized by topological chern numbers , provided that the bands are defined on a 2-torus . to that end , we supplement the bloch phase parameter in momentum space with the periodic parameter @xmath6 . this procedure somewhat resonates with recent efforts in identifying analogs of quantum hall effect in one - dimensional systems @xcite . the eigenstate - eigenvalue problem for @xmath33 now becomes @xmath34 , where @xmath35 is the bloch phase in momentum space and @xmath36 is the eigenphase of @xmath37 . for a fixed pair of @xmath6 and @xmath38 , @xmath27 eigenphases can be obtained , and scanning @xmath39 over @xmath40 forms _ extended _ floquet bands on 2-torus , so - named to distinguish them from the common bands involving only the bloch phase parameter @xmath38 . as a result @xmath41 ( @xmath42 ) becomes an index of such bands ( with @xmath43 ) . the eigenstates @xmath44 are chosen such that they are locally single - valued functions of @xmath6 and @xmath38 , and periodic functions of @xmath38 . two computational examples of the floquet bands @xmath36 are depicted in fig . [ bands ] for @xmath45 . it is seen that as the driving strength @xmath7 varies , the landscape of the bands changes . other calculations show that if @xmath25 with even @xmath27 , then there will be two ( extended ) bands touching each other . for simplicity , here we consider only odd @xmath27 so that only accidental band collisions occur . in fig . [ cherntable ] we present the chern numbers ( see its surface integral expression below and see @xcite for computational details ) for a 3-band case . it is seen that at some isolated critical values of @xmath46 , the chern numbers jump , signaling the presence of toplogical phase transitions in our ra model . indeed , the chern numbers are invariant integers with respect to smooth deformation of the bands and only change discontinuously due to band collisions . note that mainly in the context of quantum - classical correspondence in classically chaotic systems , ref . @xcite ( see also ref . @xcite ) formally studied the topological aspects of floquet bands defined on a 2-torus formed by bloch phases in both position space and momentum space . our study is much different because ( i ) this work is based on an explicit physical implementation of momentum - space periodicity , ( ii ) here the bloch phase in position space is fixed at @xmath14 in theory ( so as to obtain hofstadter s butterfly floquet spectrum ) , and ( iii ) our floquet bands are defined on a 2-torus that involves one experimental parameter , a key starting point for our theory below . we now seek the implications of the floquet - band chern numbers by considering an adiabatic cycle during which @xmath6 increases from @xmath47 to @xmath24 . we first construct an initial state of the following form , @xmath48 which represents an equal - weight superposition of all the floquet eigenstates of band @xmath41 with @xmath30 . this coherent superposition state , which can be interpreted as a wannier function in momentum space , uniformly samples all the bloch eigenstates with different values of @xmath38 ( but all with @xmath30 ) , with a profile localized in the momentum space . indeed , each eigenstate @xmath49 is infinitely extended in momentum space , but @xmath50 is normalized to unity ( localized in momentum space with gaussian - like tails in all the cases we studied ) . it is worth noting that because each eigenstate @xmath51 is defined only up to a global phase , one is free to choose an overall phase convention of @xmath52 such that the superposition state in momentum space tends to be well - localized , thus making experimental preparation of the initial state easier . such states can be highly localized so long as @xmath7 is not too large . for example , the shown superposition state in fig . 3(a ) mainly occupies one momentum eigenstate , with small weights distributed over only a few nearby components . given previous experiments where momentum superposition states in the same context were prepared @xcite , states as shown in fig . 3(a ) should be reachable in experiments . consider then an adiabatic change in @xmath6 , through a discretized protocol @xmath53 for @xmath54 , so that @xmath6 completes one adiabatic cycle at @xmath55 @xcite . assuming adiabatic following of the floquet states , the state evolved from @xmath50 [ see eq . ( [ initialstate ] ) ] should stay as a superposition state at @xmath56 , with each component still being the eigenstate of @xmath37 , with @xmath57 . that is , at @xmath58 , under adiabatic approximation the associated time - evolving state of the system becomes @xmath59 where @xmath60 is the sum of a dynamical phase and a geometrical phase accumulated by the component starting from @xmath49 . as a consequence of choosing @xmath49 to be a periodic function of @xmath38 , @xmath60 is also necessarily periodic in @xmath38 . next we evaluate @xmath61 , namely , the momentum expectation value on the state @xmath62 . to proceed we first write floquet eigenstates @xmath63 as the product of two parts using the bloch theorem , namely , @xmath64 , with @xmath65 . we then define @xmath66 . it can be shown that the @xmath27-element state @xmath67 is an eigenstate of the following @xmath68 reduced floquet matrix @xmath69 with @xmath70 . lengthy but straightforward calculations @xcite then yield a compact expression for @xmath71 , i.e. , @xmath72 . to compare this expectation value with its preceding value for @xmath73 , we consider a first - order perturbation theory so as to rewrite @xmath74 in terms of @xmath75 . specifically , to the first - order of @xmath76 , we have @xcite @xmath77 the change in momentum expectation value ( denoted by @xmath78 ) over the period from @xmath79 to @xmath80 can now be calculated to the first order of @xmath81 , @xmath82 further substituting eq . into eq . yields @xmath83 where @xmath84 is identified as the berry curvature @xmath85 , @xmath86 and @xmath87 on @xmath6 and @xmath38 all suppressed for brevity . the total change in momentum expectation value from @xmath88 to @xmath80 is denoted by @xmath89 . because @xmath90 , we have @xmath91 where we have used @xmath25 and @xmath92 is exactly the chern number of the @xmath41th floquet band , whose surface integral expression is @xmath93 @xcite . thus , eq . ( [ finaleq ] ) reveals that the net change in the momentum expectation value over one adiabatic cycle of @xmath6 ( starting from state @xmath50 ) is quantized : it should be proportional to the chern number of the @xmath41th extended floquet band . this is our central theoretical result . c + + it is necessary to numerically verify our theoretical insights above . detailed results are shown in fig . [ transport_numbers ] , again for the case of @xmath94 , with the adiabatic cycle lasting for @xmath95 periods ( in this case , see also fig . 3(b ) for the final momentum space profile ) or @xmath96 periods . first of all , apart from the regime of @xmath46 values near the critical point @xmath97 ( see fig . 2 ) , our numerical values of @xmath98 almost perfectly match the chern numbers . this is the case before or after the jumps of the chern numbers . the insets of both panels depict how @xmath99 builds up with time and eventually reaches integer values that match the chern numbers . in addition , we have checked numerically that if we repeat the adiabatic cycle , then the same quantized increase in momentum expectation value is obtained @xcite . we are thus witnessing a clear quantization effect in acceleration as an outcome of floquet band topology . note however , in the vicinity of phase transition points , e.g. , @xmath100 , momentum space transport is no longer quantized . this is because if a topological phase transition is about to occur , then the associated band gaps are not large enough to guarantee adiabaticity . supporting this understanding , a comparison between fig . [ transport_numbers ] ( b ) and fig . [ transport_numbers](a ) shows that a longer adiabatic cycle indeed significantly narrows down the non - quantization window . our numerical data suggests that at least for the 3-band case here , if the driving field strength @xmath46 is far away from the phase transition points , then only 50 - 100 kicking periods ( depending on @xmath46 ) will be needed to observe quantized acceleration . this is experimentally motivating , because floquet state manipulation itself is a topic of much theoretical interest @xcite . the robustness of this quantization against perturbations is also examined in appendix @xcite . in short , for the 3-band case above , the quantization effect can tolerate about 0.5% uncertainty in @xmath31 and a nonzero @xmath15 around 0.01 , which should be achievable in light of previous experiments @xcite . in conclusion , based on extended floquet bands , we have exposed topological phase transitions in driven quantum systems and demonstrated how quantized adiabatic transport in momentum space may emerge from floquet band topology . numerical results based on a cold - atom - based dynamical model suggest that future experimental verification of our results is possible in terms of initial state preparation , adiabatic cycle implementation , and the robustness of quantization . we thank jiao wang for providing some useful codes during the early stage of this work . we thank adam zaman chaudhry , wayne lawton , and lakhaphat lin aigu for helpful discussions . j.g . acknowledges funding from academic research fund tier i , ministry of education , singapore ( grant no . r-144 - 000 - 276 - 112 ) . 99 d. r. grempel , r. e. prange , and s. fishman , * 29 * , 1639 ( 1984 ) . g. casati and b.v . chirikov , _ quantum chaos : between order and disorder _ ( cambridge university press , new york , 1995 ) . j. chab , g. lemari , b. grmaud , d. delande , p. szriftgiser , and j. c. garreau , * 101 * , 255702 ( 2008 ) ; g. lemari , h. lignier , d. delande , p. szriftgiser , and j. c. garreau , * 105 * , 090601 ( 2010 ) ; m. lopez , j. clement , p. szriftgiser , j. c. garreau , d. delande , * 108 * , 095701 ( 2012 ) . j. b. gong , and p. brumer , * 70 * , 016202 ( 2004 ) ; j. b. gong , and p. brumer , * 97 * , 240602 ( 2006 ) . e. lundh and m. wallin , * 94 * , 110603 ( 2005 ) . i. dana , v. ramareddy , i. talukdar , and g. s. summy , * 100 * , 024103 ( 2008 ) ; m. sadgrove , h. munekazu , t. sekimura , and k. nakagawa , * 99 * , 043002 ( 2007 ) . floquet bands ( traditionally due to periodicity in position space ) are already of wide interest , see , for example , t. kitagawa , e. berg , m. s. rudner , and e. demler , * 82 * , 235114 ( 2010 ) . see also n. h. lindner , g. refael , and v. galitski , nature physics , * 7 * , 490 ( 2011 ) for a proposal of floquet topological insulators . d. j. thouless , * 27 * , 6083 ( 1983 ) . d. w. hone , r. ketzmerick , and w. kohn , * 56 * , 4045 ( 1997 ) . a. dranov , j. kellendonk , and r. seiler , j. math . phys . * 39 * , 1340 ( 1998 ) ; a. tanaka and m. miyamoto , * 98 * , 160407 ( 2007 ) ; a. tanaka , j. phys . soc . japan * 80 * , 125002 ( 2011 ) . j. wang and j. b. gong , * 78 * , 036219 ( 2008 ) . h. jones , m. m. stocklin , g. hur , and t. s. monteiro , * 93 * , 223002 ( 2004 ) ; c. e. creffield , g. hur , and t. s. monteiro , * 96 * , 024103 ( 2006 ) . c. ryu , m. f. andersen , a. vaziri , m. b. darcy , j. m. grossman , k. helmerson , and w. d. phillips , * 96 * , 160403 ( 2006 ) ; in this experiment , the uncertainty in quasi - momentum @xmath15 is 0.002 , already smaller than what we need . i. talukdar , r. shrestha , and g. s. summy , * 105 * , 054103 ( 2010 ) . see appendices for details of chern number calculations , more momentum - space distributions , effects of a nonzero @xmath15 and uncertainty in @xmath101 , experimental noise effects , detailed theoretical derivations , as well as chern number calculations for a 7-band case . f. l. moore , j. c. robinson , c. f. bharucha , b. sundaram , and m. g. raizen , * 75 * , 4598 ( 1995 ) . j. f. kanem , s. maneshi , m. partlow , m. spanner , and a. m. steinberg , * 98 * , 083004 ( 2007 ) . a. ullah and m. d. hoogerland , * 83 * , 046218 ( 2011 ) . j. wang and j. b. gong , * 77 * , 031405 ( 2008 ) . d. r. hofstadter , * 14 * , 2239 ( 1976 ) . j. p. dalhaus , j. m. edge , j. tworzydo , and c. w. j. beenakker , * 84 * , 115133 ( 2011 ) ; l. j. lang , x. cai , and s. chen , , 220401 ( 2012 ) ; f. mei , s. l. zhu , z. m. zhang , c. h. oh , and n. goldman , * 85 * , 013638 ( 2012 ) . y. e. kraus , y. lahini , z. ringel , m. verbin , and o. zilberberg , arxiv:1109.5983v2 . p. leboeuf , j. kurchan , m. feingold , and d. p. arovas , * 65 * , 3076 ( 1990 ) . i. dana , * 52 * , 466 ( 1995 ) ; i. dana , m. feingold , and m. wilkinson , * 81 * , 3124 ( 1998 ) . exact @xmath6 values at each step are not essential . on the other hand , see m. sadgrove and k. nakagawa , rev . sci . instrum . * 82 * , 113104 ( 2011 ) for a recent demonstration of optical lattice phase control within 7.2% error of the standing wave period , on a time scale of 28 @xmath102 . s. fishman , i. guarneri , and l. rebuzzini , j. stat . phys . * 110 * , 911 ( 2003 ) .
the last two decades have been witness to intense activities in experimental and theoretical study of correlation properties of interacting quantum systems . the problem is still challenging and open in the quantum many - body systems which are non - integrable . the task becomes simplified for integrable systems like one dimensional uniform bose gas described by the lieb - liniger ( ll ) model [ 1,2 ] , which assumes that particles interact via a @xmath0-function repulsive potential . since the experimental observation of bose einstein condensate ( bec ) in ultracold trapped alkali atomic vapors , a lot of theoretical and experimental work has been done to study its correlation properties [ 3 - 12 ] . in the recent experimental situation , it is easy to achieve a quasi - one - dimensional strongly interacting degenerate bose gas in a highly anisotropic trap [ 13 - 15 ] which is correctly described by the lieb - liniger model [ 16 ] . the complementary case is the weakly interacting ( via finite range forces ) trapped bose gas . although it is commonly believed that the gross pitaveskii ( gp ) equation is adequate for weakly interacting bose gases , but a more rigorous and accurate many - body treatment , incorporating realistic interatomic interactions and interatomic correlations , is crucial for studying correlation properties in realistic condensates . the first motivation of the present work is to investigate the importance of finite size and trapping effects on the ground state correlation properties of ultracold atomic becs . the second motivation of our study is to assess the validity of shape - independent approximation in the correlation properties . dilute becs are known to possess shape - independent ( si ) property which is frequently described by the @xmath0-function potential , whose strength is proportional to the @xmath1-wave scattering length @xmath2 . in the mean - field description , the effective interaction is determined by the factor @xmath3 ( where @xmath4 is expressed in units of oscillator length of the trap ) , whereas in the full many - body description we solve the many - body schrdinger equation for different two - body potentials that generate identical @xmath5 . thus , our present calculation serves as a stringent test of the si approximation and verifies whether the long range correlation at all affects the correlation properties or not . inclusion of all possible two - body correlations in our _ ab initio _ many - body method correctly calculates correlation effects at zero temperature . use of the realistic van der waals interatomic potential with a long attractive tail provides the realistic aspects of the correlation function and its momentum distribution . specially the effect of short range repulsion in the interatomic potential is expected to be reflected in the pair correlation function . + the paper is organized as follows . in the next section , we briefly review our theoretical approach . in section iii , we present the results of our calculation of one - body density , pair distribution function and their momentum distributions . finally , in section iv we draw our conclusions we adopt the potential harmonics expansion method together with a short range correlation function ( cphem ) , which has already been established as a successful and useful technique for investigating bec with realistic two - body interactions . in the following , we briefly describe the technique . interested readers can find the details in refs . + we start with the many - body schrdinger equation of @xmath6 spinless bosons trapped in an isotropic harmonic oscillator potential of frequency @xmath7 at zero temperature . the center of mass motion can be separated by introducing the center of mass vector ( @xmath8 ) and @xmath9 jacobi vectors @xmath10 defined as @xcite @xmath11 where @xmath12 is the position vector of the @xmath13-th particle . the relative motion of the system is described by @xmath14 \psi(\vec{\zeta}_{1 } , ... , \vec{\zeta}_{n } ) = 0 , \label{se_rel_motion}\end{aligned}\ ] ] where @xmath15 and @xmath16 are respectively the trapping and pair - wise interaction potentials , expressed in terms of the jacobi vectors . the energy of the relative motion is @xmath17 . next hyperspherical variables are introduced by defining a ` hyperradius ' @xmath18^{\frac{1}{2}},\ ] ] and a set of @xmath19 ` hyperangles ' , consisting of @xmath20 polar angles of @xmath21 jacobi vectors and @xmath22 angles defining their _ relative lengths _ @xcite . in the hyperspherical harmonics expansion method ( hhem ) the @xmath21-body schrdinger equation ( [ se_rel_motion ] ) is solved by expanding @xmath23 in the complete set of hyperspherical harmonics ( hh ) , which are the eigenfunctions of the @xmath21-dimensional laplace operator ( analogous to the spherical harmonics in three dimension ) @xcite . substitution of this in eq . ( [ se_rel_motion ] ) and projection on a particular hh results in a set of coupled differential equations . however imposition of symmetry of the wave function and calculation of the matrix elements are such formidable tasks that a practical solution for @xmath24 is nearly impossible . moreover due to the fact that the degeneracy of the hh basis increases very rapidly with the increase in the grand orbital quantum number @xmath25 @xcite , a convergent calculation using hhem with _ the full _ hh basis is extremely computer intensive . this is the price one pays for keeping _ all many - body correlations_. on the other hand , a typical experimental bec is designed to be extremely dilute to eliminate the possibility of molecule formation through three - body collisions . hence we assume that the probability of three and more particles to come within the range of interatomic interaction is negligible and the effect of the two - body correlations will be adequate for the full many - body wave function . moreover , only two - body interactions are relevant . therefore , @xmath23 can be expressed in terms of faddeev components @xmath26 of the @xmath27 interacting pair @xcite @xmath28 since only two - body correlations are important , @xmath26 is a function of the interacting - pair separation @xmath29 and @xmath30 only . hence it can be expanded in the subset of hh needed for the expansion of the two - body interaction @xmath31 . this subset is called the potential harmonics ( ph ) basis @xcite . this results in a dramatic simplification in the analytic and computation works and it has been used upto 14000 bosons in the trap @xcite . in this procedure , a realistic interatomic interaction can be used . such an interaction has a strongly repulsive core . hence @xmath26 must be vanishingly small for small values of @xmath32 . but the first ( @xmath33 ) term of the ph basis is a constant @xcite . hence to reproduce the short - range behaviour of @xmath26 correctly , a large number of ph basis functions is needed , which slows down the rate of convergence considerably . hence , we introduce a ` short - range correlation function ' in the ph expansion basis @xcite . in the zero temperature bec , the kinetic energy of the interacting pair is practically zero . therefore , the two - body collision process is determined entirely by the @xmath1-wave scattering length ( @xmath4 ) . the small @xmath32 behavior of @xmath26 will be that of the zero - energy wave function @xmath34 of the pair interacting via @xmath35 . hence we introduce @xmath34 as a short - range correlation function in the ph expansion @xcite . we have checked explicitly that this improves the rate of convergence dramatically . in order that the small @xmath32 behavior of @xmath36 corresponds to the correct two - body interaction appropriate for the experimental @xmath4 , we adjust the very short - range behavior of the interatomic interaction ( in the case of the van der waals potential used by us , it will be the hard - core radius @xmath37 ) so that @xmath34 asymptotically becomes proportional to @xmath38 @xcite . substitution of this expansion of @xmath23 in eq . ( [ se_rel_motion ] ) and projection on a particular ph gives rise to a set of coupled differential equations in @xmath30 @xcite @xmath39 u_{kl}(r ) \\ & + & \sum_{k^{\prime}}f_{kl}v_{kk^{\prime}}(r)f_{k'l } u_{k^{\prime}l}(r ) = 0 , \nonumber \label{finalcde}\end{aligned}\ ] ] where @xmath40 , @xmath41 , @xmath42 and @xmath43 . here , @xmath44 is the orbital angular momentum of the system ( assumed to be contributed by the interacting pair only ) and @xmath45 is the coefficient of expansion of @xmath36 in the correlated ph basis , while @xmath46 represents the overlap of the ph corresponding to the ( @xmath47)-partition with the sum of phs of all partitions @xcite . expressions for the potential matrix element @xmath48 and @xmath46 can be found in ref . ( 5 ) is solved by hyperspherical adiabatic approximation ( haa ) @xcite , for which we assume that the hyperangular motion is much faster than the hyperradial motion . consequently , the former can be separated adiabatically and solved to obtain an effective potential as a parametric function of @xmath30 , in which the hyperradial motion takes place . the hyperangular motion is effectively solved by diagonalizing the potential matrix @xmath48 together with the hyper - centrifugal potential of eq . ( 5 ) to get the lowest eigenvalue @xmath49 [ corresponding eigen column vector being @xmath50 , which is the effective potential for the hyperradial motion . finally , the adiabatically separated hyperradial equation @xmath51\zeta_0(r)=0.\ ] ] is solved subject to appropriate boundary conditions to get @xmath17 and the hyperradial wave function @xmath52 . the many - body wave function can be constructed in terms of @xmath52 and @xmath53 @xcite . now in order to study the correlation properties of the interacting bose gas we choose a realistic interatomic potential having a strong repulsive core at a small separation and an attractive tail at large atomic separations . this is approximately represented by the van der waals potential with a hard core of radius @xmath54 and a @xmath55 attractive tail , _ viz . _ @xmath35 = @xmath56 , for @xmath57 and @xmath58 for @xmath59 . the effective interaction is characterized by the @xmath1-wave scattering length ( @xmath5 ) , which depends strongly on @xmath54 @xcite . usually the potentials are chosen to be purely attractive or purely repulsive according to whether @xmath5 is negative or positive respectively . in our many - body calculation , we solve the zero - energy two - body schrdinger equation with @xmath35 to obtain @xmath5 @xcite . the value of @xmath54 is adjusted so that @xmath5 has the values corresponding to the jila experiments @xcite . a typical value of @xmath37 is of the order of @xmath60 o.u . in atomic units this is a few tens of bohr , which is larger than the atomic radius . note that @xmath37 is expected to be larger than the atomic radius , as the van der waals potential with the hard core of radius @xmath37 is the _ effective potential _ which produces the correct experimental _ zero - energy _ scattering cross - section ( given by @xmath4 ) . + + taking the fourier transform of @xmath78 , we obtain the one - body momentum distribution . one - body density is an abstract concept . its fourier transform gives the experimentally measurable quantity , the momentum distribution . in fig . 3 , we plot our results for repulsive interactions . it evolves from a gaussian in the noninteracting limit to a curve having a sharper peak , as the net interaction increases . the peak at @xmath79 becomes more pronounced with increase in effective repulsion , whereas for weak interaction it develops a long - range tail in the momentum space . the momentum is being redistributed to higher @xmath80 values . the width of the low - momentum peak for @xmath81 o.u . and @xmath82 is about @xmath83 @xmath84m@xmath85 . we have remarked earlier that with recent progress in creating atomic clouds with large dipole moment , interest has been shifted to longer range interaction instead of taking only contact interaction . thus , even in the low - density limit the use of a realistic interatomic interaction potential in the many - body calculation has been emphasized by several authors [ 27 - 29 ] . the van der waals potential is an ideal choice as it properly takes care of the effect of realistic dipole - dipole interaction . the strength of the van der waals interaction @xmath86 changes widely from a small value for h atoms to a high value for cs atoms . hence , to see how the one - body density is affected by the strength of the long - range tail of the two - body potential , we select three different values of @xmath86 in addition to the actual experimental value one below and two above . these are ( in o.u . ) : @xmath87 , @xmath88 , @xmath89 , @xmath90 , the second one being the experimental value . for each value of @xmath86 , we calculate corresponding @xmath54 as before , such that @xmath5 has the experimental value 0.00433 o.u . for each set of @xmath91 , we solve the many - body schrdinger equation as before . calculated one - body density for different sets are shown in fig . 4 for 10000 bosons . we observe that the one - body density of the inhomogeneous gas for different two - body potentials are almost indistinguishble ; i.e. independent of the shape of @xmath35 . our numerical calculations confirm that one - body density is absolutely determined ( within numerical errors ) by the parameter @xmath5 only . this is in contrast with ref . @xcite , where ground state energy of the condensate was found to depend on the shape of the potential . thus the one - body density does not depend strongly on the shape of the interaction potential , while the ground state energy does . + an explanation of the above observation is as follows . the total condensate energy depends on the minimum of the well , in addition to the actual shape of the lowest eigen potential . in table 1 , we present the dependence of the position ( @xmath92 ) and the value ( @xmath93 ) of the minimum of lowest eigen potential for different values of @xmath94 parameter , corresponding to the same @xmath4 . + table 1 . dependence of @xmath92 and @xmath93 on @xmath94 ( corresponding to @xmath95 ) for @xmath82 atoms ( all quantities are expressed in o.u . ) . + [ cols="^,^,^",options="header " , ] we observe that , as @xmath94 increases , @xmath93 decreases appreciably , while position of the minimum of effective many - body potential changes by a small amount . in the next sub - section , we will see that the shape of the @xmath49 curve remains practically unchanged . as the total condensate energy depends on the depth of the potential , as well as on its shape ( stiffness ) , we observe that the ground state energy decreases gradually with increase in @xmath94 @xcite . thus the shape independence hypothesis is violated for the total energy . however , the one - body density distribution is given by the many - body wave function , which is independent of the minimum of the potential , but depends only on its shape . hence it remains unchanged with change in @xmath94 parameter and shape independence of the one - body density profile is satisfied . + another key quantity is the pair distribution function @xmath96 , which determines the probability of finding the @xmath27-pair of particles at a relative separation @xmath32 . the study of pair correlation is important , since the interatomic interactions play a crucial role as there are two competing interaction length scales . when the atoms try to form clusters , a strong very short range repulsion in atomic interaction comes into play . as atoms repel each other strongly at very small separations it is impossible to get some non - vanishing value of @xmath96 at @xmath32 = 0 . we calculate it as @xmath97 where @xmath23 is the many - body wavefunction and the integral over the hypervolume @xmath98 excludes integration over @xmath99 . again after a lengthy calculation we can put it in a closed form given by @xmath100 we plot @xmath96 in fig . 5 for the attractive interaction with different particle numbers . as mentioned earlier , pair correlation vanishes as @xmath32 @xmath101 0 , due to the strong interatomic repulsion and it can not extend beyond the size of the condensate . hence @xmath102 is peaked at some intermediate value of @xmath32 . for weak interactions ( small @xmath77 ) , the correlation length is large . however as the effective attractive interaction increases , two interacting particles come closer and the pair - correlation length decreases ; the pair distribution function becomes sharply peaked . it indicates stronger pair correlation in the system , in agreement with expectations . however our results are at variance with those obtained in ll model @xcite , which describes uniform systems with no confinement . consequently pair correlation function approaches its maximum asymptotically , whereas in our confined three dimensional case , it vanishes asymptotically . when @xmath6 is increased to 2470 , which is very close to the critical number ( @xmath103 2475 ) , the effect is quite prominant . as the width of the curves decrease with increase in effective attractive interaction , the interacting pair becomes more localized . this manifests the possiblity of clustering due to large two - body interaction and three - body recombination , if the attraction becomes strong enough . this will lead to an eventual collapse of the attractive condensate , when the pair - correlation length will drastically reduce to a very small value and the condensate will be destroyed . + two - particle correlation is directly related with pair inelastic processes and can be directly measured in photoassociation in interatomic collisions . as the attractive interaction increases , the gas becomes highly correlated and different inelastic processes may take place . the opposite case is the repulsive bose gas where the pair inelastic process will be suppressed with increase in interaction . for completeness , in fig . 6 we plot the relative momentum distribution obtained as the fourier transform of @xmath96 for various particle numbers for @xmath69rb condensate in the jila trap , with @xmath5 = @xmath104 o.u . ( @xmath105 a.u . ) . the width of the curve decreases and exhibits a sharp peak at @xmath106 , whose height increases with increasing number of particles . + lastly to visualize the effect of different strength parameter @xmath86 on the pair - correlation , we plot @xmath96 against @xmath32 for the previous sets of potential parameters in fig . 7 . we find that different interatomic potential shapes corresponding to the same @xmath5 produce identical correlation properties , which indicates that the calculated property is independent of the shape of the two - body potential . the fact that the ground state energy is dependent on the shape of the two - body potential @xcite whereas the one - body density and the pair distribution function are not , can be attributed to the fact that the many - body wave function remains invariant irrespective of the variation of the shape of the two - body potential due to variation of @xmath107 for the same value of @xmath4 . note that the condensate wave function in the hyperradial space @xmath108 is obtained by solving the hyperradial equation ( 6 ) in the effective many - body potential @xmath109 . the shape and stiffness of this effective potential remain unchanged for the same value of @xmath5 , irrespective of the variation of the shape of the two - body potential . however , as discussed earlier , the position and value of the minimum of the effective many - body potential change with the variation of the parameter @xmath86 . to demonstrate this , we plot the effective many - body potential @xmath109 as a function of @xmath30 for various values of @xmath86 in fig . the curves have been shifted both vertically and horizontally in order that the minima coincide ( both in position and value ) . one observes that all the curves overlap completely ( within numerical errors ) showing that the shape and stiffness of the effective potential remain invariant . hence the kinetic energy and the wave function remain unchanged , while potential energy and the total energy change with change of @xmath94 . by direct numerical calculation , we have checked that the change of kinetic energy is very small compared with that of the total energy , as @xmath94 varies . therefore the ground state energy is shape dependent , whereas the one - body density and pair distribution function are not . in conclusion , the present work focuses on the one - body density and pair - correlation aspects of the zero temperature weakly interacting trapped bose gas . the use of correlated many - body approach takes care of the effect of finite - size , where quantum fluctuation is important and gives a realistic picture of correlation properties . due to the use of a realistic interatomic interaction and consideration of finite number of atoms in the trap , our results deviate from the earlier results @xcite , but exhibit realistic aspects which are relevant to experiments . our calculation is performed for a two - body potential ( van der waals potential ) , whose parameters can be adjusted to give both positive and negative scattering lengths , for rb atoms in the jila trap . thus , our results are realistic and can be experimentally verified in future . our calculations also verify the validity of the shape - independent approximation in dilute becs for one - body density and pair - distribution functions . this is in contrast with the earlier observed shape dependence for ground state energies @xcite . + we would like to acknowledge helpful discussion with l. salasnich , s.r . jain and g. mussardo . this work has been funded by the department of science and technology ( dst ) , india under grant number sr / s2/cmp/0059(2007 ) . bc thanks the international center for theoretical physics ( ictp ) , trieste for hospitality and financial support , where a part of the work was done . tkd acknowledges the university grants commission ( ugc ) , india for the emeritus fellowship . ab acknowledges the ugc , india for the research fellowship in sciences for meritorious students and the council of scientific and industrial research ( csir ) , india for a senior research fellowship . + references e. h. lieb and w. liniger , phys . rev . * 130 * , 1605 ( 1963 ) . e. h. lieb , phys . rev . * 130 * , 1616 ( 1963 ) . k. sakman _ et al _ , phys . a * 78 * , 023615 ( 2008 ) . m. naraschewski and r.j.glauber , phys . a * 59 * , 4595 ( 1999 ) . l. e. sadler _ et al _ , phys . lett . * 98 * , 110401 ( 2007 ) . s. ritter _ et al _ , phys . lett . * 98 * , 090402 ( 2007 ) . i. bloch , t.w . hnsch and t.esslinger , nature(london ) , * 403 * , 166 ( 2000 ) . g. e. astrakharchik and s. giorgini , phys . a * 68 * , 031602(r ) ( 2003 ) . s.zllner , h.d.meyer and p.schmelcher , phys . lett . * 100 * , 040401 ( 2008 ) . r. pezer and h. buljan , phys . lett . * 98 * , 240403 ( 2007 ) . k. v. kheruntsyan _ et al _ , phys . * 91 * , 040403 ( 2003 ) . d. m. gangardt and g. v. shlyapnikov , phys . 90 * , 010401 ( 2003 ) . m. greiner _ et al _ , phys . lett . * 87 * , 160405 ( 2001 ) . f. schreck _ et al _ , phys . lett . * 87 * , 080403 ( 2001 ) . a. grlitz _ et al _ , phys . lett . * 87 * , 130402 ( 2001 ) . d. m. gangardt and g.v . shlyapnikov , new journal of phys . * 5 * , 79 ( 2003 ) . t.k.das and b.chakrabarti , phys a * 70 * , 063601 , ( 2004 ) . t.k.das , s. canuto , a. kundu and b. chakrabarti , phys . a * 75 * , 042705 , ( 2007 ) . t. k.das , a. kundu , s. canuto and b. chakrabarti , phys . a * 373 * , 258 ( 2009 ) . j.l.ballot and m.fabre de la ripelle , ann . ( n.y . ) * 127 * , 62 , ( 1980 ) . m.fabre de la ripelle , ann . ( n.y . ) * 147 * , 281 , ( 1983 ) . c. d. lin phys . rep . * 257 * , 1 ( 1995 ) . c.j . pethick and h. smith , bose - einstein condensation in dilute gases , cambridge university press , cambridge , ( 2001 ) . t.k.das , h.t.coelho and m.fabre de la ripelle , phys . c * 26 * , 2281 , ( 1982 ) . m. h. anderson _ et al _ , science * 269 * , 198 ( 1995 ) . roberts et al , phys . lett . * 86 * , 4211 , ( 2001 ) . b. gao , j. phys . b * 36 * , 2111 ( 2003 ) . s. geltman , j. phys . b * 37 * , 315 ( 2004 ) . b. chakrabarti and t. k. das , phys . a * 78 * , 063608 ( 2008 ) .	we study the ground state pair - correlation properties of a weakly interacting trapped bose gas in three dimension by using a correlated many - body method . use of the van der waals interaction potential and an external trapping potential shows realistic features . we also test the validity of shape - independent approximation in the calculation of correlation properties .
the physically relevant degrees of freedom of a large isolated system initially out of equilibrium , are expected to relax to their thermal equilibrium values . obviously , no relaxation behavior can be observed for the complete quantum state of the system but interesting degrees of freedom can evolve irreversibly in the limit of a large system . this issue of thermalisation in isolated quantum systems has been principally investigated by considering composite systems consisting of two distinguishable subsystems of very different sizes . for such systems , the smaller subsystem relaxes to thermal equilibrium if the larger one is assumed to be in an equilibrium mixed state @xcite . recently , it has been shown that a priori thermal averaging is not essential . the small subsystem can thermalise whereas the whole isolated system is in a pure quantum state @xcite . an isolated system of identical particles can not be divided into distinguishable subsystems and the relevant degrees of freedom are , in this case , the n - particle reduced distribution functions such as the particle number density . relaxation behaviors have been obtained for both integrable @xcite and nonintegrable systems @xcite . in these studies , the system is supposed to be initially in the ground state of a given hamiltonian but its subsequent time evolution is governed by a different hamiltonian . for example , a gas is considered to be initially confined in a subvolume of a larger accessible space . the single - particle distributions were always found to relax to asymptotic profiles but which are , intriguingly , not all well described by an equilibrium ensemble . to better apprehend the thermalisation process in isolated many - body systems , it is necessary not to be limited to ground states and to be able to treat also finite initial temperatures . as we are concerned with isolated systems , such an initial condition must be represented by a microcanonical mixed state , according to the postulate of equal a priori probabilities , or , equivalently , by a pure state of macroscopically well - defined energy @xcite . in ref.@xcite , the time - dependent particle density of a perfect quantum gas , initially in such a state , has been determined . it was found that this density relaxes from a thermal profile to a non - thermal one . in view of these results , one may wonder how two quantum systems , initially independent of each other but at the same temperature , evolve if they are brought into thermal contact . do their relevant degrees of freedom remain at equilibrium once they are coupled ? in this paper , we address this issue by considering a free bosonic field in two one - dimensional cavities coupled to each other by one of their ends . free fields are commonly used to describe the large number of environmental degrees of freedom which give rise to dissipation in a smaller system @xcite . here , we are interested in two systems of comparable size , neither of them acts as a heat bath for the other one . in section [ m ] , we present our model hamiltonian of two quantum systems in thermal contact . this model allows to describe from perfectly coupled to completely uncoupled systems . we show that the considered coupling between the two cavities is thermodynamically small by evaluating the equilibrium total energy and correlation functions . in section [ scc ] , we determine how the energy contents of the cavities and the field - field correlations evolve in response to a sudden change of the coupling between the two cavities . our model is simple enough to obtain exact results for arbitrary values of the coupling strength . in section [ cpg ] , we discuss why , though the expressions are formally similar , relaxation behaviors are obtained for a perfect gas but not for the system studied here . finally , in the last section , we summarize our results and mention a few open questions raised by our work . to describe two coupled one - dimensional bosonic systems characterised by length @xmath0 and velocity @xmath1 , we start from the hamiltonian @xmath2 \label{hp}\ ] ] where the fields @xmath3 and @xmath4 are canonically conjugate to each other . the position - dependent refractive index @xmath5 is equal to @xmath6 for @xmath7 and to @xmath8 elsewhere . diverse physical systems can be represented by this hamiltonian . the fields @xmath3 and @xmath9 can be interpreted as the electric and magnetic components of the electromagnetic field , or as the charge and current distributions of an lc transmission line @xcite . we write these fields as @xmath10 where the operators @xmath11 and @xmath12 satisfy the commutation relations @xmath13=[a_q , a_{q'}]=0 $ ] and @xmath14=\delta_{q{q'}}$ ] , and @xmath15 is a cut - off length . throughout this paper , we use units in which @xmath16 . the eigenmodes @xmath17 of @xmath18 are solutions of the differential equation @xmath19 $ ] with the boundary conditions @xmath20 . the odd @xmath21 are proportional to @xmath22 $ ] for @xmath23 , where the wavenumber @xmath24 obeys @xmath25^{-1}=n\tan(2nqd ) . \label{eqq0}\ ] ] the even eigenmodes are proportional to @xmath26 $ ] for @xmath23 , with @xmath24 given by @xmath27=-n\tan(2nqd)$ ] . to obtain the hamiltonian of two cavities coupled to each other by one of their ends , we consider the limits @xmath28 and @xmath29 with the length @xmath30 kept fixed . in this regime , equation simplifies to @xmath31 which can be rewritten as @xmath32 this second form will be useful in the following . for the even eigenmodes , the wavenumbers become @xmath33 where @xmath34 is a positive integer . it can then be shown , using and , that the hamiltonian turns into @xmath35 where @xmath36 , \label{hl}\ ] ] @xmath37 is given by this expression with @xmath38 and @xmath39 replaced by @xmath39 and @xmath0 , respectively , and @xmath40 ^ 2 . \label{hc}\ ] ] in the following , we study the hamiltonian which allows to describe from perfectly coupled to completely uncoupled cavities . its eigenmodes @xmath17 obey @xmath41/d$ ] which leads to even @xmath42 independent of the coupling characteristic length @xmath43 . the odd eigenmodes read @xmath44 \label{oe}\ ] ] where @xmath45^{-1/2}$ ] and @xmath24 is a positive solution of . in the large @xmath43 limit , the two cavities decouple from each other as can be seen from the fact that the boundary conditions at the ends @xmath46 and @xmath47 become similar to that at the ends @xmath38 and @xmath0 . in this limit , the solutions @xmath24 of the equation are the multiples of @xmath48 and the corresponding eigenmodes @xmath49 simplify to @xmath50 for @xmath51 and to @xmath52 for the solution @xmath53 . in the opposite limit , @xmath54 , the odd eigenmodes are given by @xmath55 with the wavenumbers @xmath56 where @xmath57 . the transmission at @xmath58 is perfect in this case . as a function of length @xmath0 for @xmath59 and different values of @xmath43 . @xmath60 reaches a maximum at @xmath61 . the corresponding curve is indistinguishable from that obtained for @xmath62 . the results for @xmath63 and @xmath64 are well - described by the large and vanishing @xmath43 approximations and , respectively . [ fig : e],scaledwidth=65.0% ] the coupling is thermodynamically small , i.e. , its contributions to the thermodynamic functions of the complete system described by are negligible in the thermodynamic limit of large @xmath0 and fixed energy density . as an example , we consider the average energy @xmath65/\mathrm{tr}[\exp(-h / t)]$ ] at temperature @xmath66 . diagonalizing the hamiltonian with the help of the expansions and gives @xmath67 as there is only one solution to equation in a given interval @xmath68 $ ] where @xmath57 , the above sum over @xmath24 can be approximated by an integral , which leads to @xmath69 the first two extensive terms stem from the integral approximation , see [ aa ] . the energy @xmath60 is the difference between this approximation and the sum and is thus expected to be negligible in the limit @xmath70 . for large and vanishing @xmath43 , the wavenumbers @xmath24 are equally spaced and it is then possible to evaluate @xmath60 , see [ aa ] . we find @xmath71 note that the energy @xmath72 for large @xmath43 is not exactly twice the energy of an isolated cavity of length @xmath0 , the coupling between the two cavities contributes an energy @xmath73 . in the limit @xmath70 , the energy reaches @xmath73 and vanishes . for an arbitrary characteristic coupling length @xmath43 , @xmath74 converges also to a finite value in this limit , as shown by fig . [ fig : e ] , and hence the energy of the complete system @xmath75 is essentially independent of @xmath43 . we remark that , since the solutions of are @xmath76 where @xmath34 is a positive integer , for large @xmath43 , @xmath60 first increases with decreasing @xmath43 . the expressions and are valid for any @xmath77 . we see that , for @xmath78 , the contribution of the coupling can not be neglected as @xmath79 and @xmath80 diverge differently in this limit . we obtain this diverging behavior of the energy @xmath60 for other values of @xmath43 as well , see fig . [ fig : e ] . as a function of @xmath81 for @xmath82 , @xmath83 and @xmath84 , and @xmath85 , @xmath0 and @xmath86 . the dotted line corresponds to the thermodynamic limit expression . for the above parameters , the agreement with this approximation is excellent for @xmath87 . [ fig : correq],scaledwidth=65.0% ] it is also instructive to consider the thermal two - point correlation functions of the system , for example @xmath88 where @xmath34 runs over all integers and the second sum over the negative and positive solutions of . for vanishing and large @xmath43 , this expression can be simplified by using the relation and the function . we find , for @xmath54 and @xmath89 , respectively , @xmath90 \label{pipi0 } \\ \langle \pi(x ) \pi(x ' ) \rangle^\infty_t & = & -\frac{\pi t^2}{2c } \theta(xx ' ) \sum_{p,\epsilon } \sinh^{-2 } \left[\frac{\pi t}{c}\big ( x+\epsilon x'+2pl \big ) \right ] \label{pipiinf } \end{aligned}\ ] ] where @xmath34 runs over @xmath91 and @xmath92 over @xmath93 . for @xmath70 , the only terms which contribute significantly to the sum are @xmath94 , @xmath95 and @xmath96 . the last two are important only for @xmath97 in the vicinity of @xmath98 and @xmath99 . the vanishing of for @xmath100 shows that there is no correlation between the two cavities in the large @xmath43 limit . the comparison of and reveals that , in the large @xmath43 case , the thermal correlation function is identical to that of two isolated cavities of length @xmath0 . consequently , in these two limiting cases , @xmath101 \label{corrttl}\ ] ] for @xmath102 and @xmath103 not too close to the cavities ends . more precisely , @xmath104 deviates notably from this approximate expression if @xmath102 or @xmath103 is at a distance smaller than @xmath105 from @xmath38 , @xmath39 or @xmath0 . for other values of @xmath43 , the field - field correlations are also well described , in the large @xmath0 limit , by . this can be seen as follows . for @xmath106 not too close to the ends @xmath38 , @xmath39 and @xmath0 , the terms @xmath107 $ ] and @xmath108 $ ] in vary a lot from one value of @xmath24 to the next and hence give vanishing contributions . moreover , the normalisation factor @xmath109 is close to @xmath110 except possibly for the first few roots of , and this equation has only one solution in a given interval @xmath68 $ ] where @xmath57 . as a result , the two remaining sums in are accurately approximated by the same integral which leads to . the thermal correlation function differs significantly from this approximation only for @xmath102 and @xmath103 close to the cavities ends as shown in fig . [ fig : correq ] . for the field @xmath111 , we obtain the expression @xmath112 where @xmath34 runs over @xmath91 and the second sum over the solutions of . for @xmath54 and @xmath89 , this correlation function is given by and , respectively , with an extra factor @xmath113 in the summand . for other values of @xmath43 , the above arguments can be used to simplify . thus , the approximate expression applies to the field @xmath111 as well . for the correlations between the fields @xmath3 and @xmath9 , the completeness of the basis @xmath114 implies , for any @xmath43 , @xmath115 . in summary , in the thermodynamic limit , the thermal correlation functions are essentially independent of the characteristic coupling length @xmath43 . we consider here that the system is initially at equilibrium with temperature @xmath66 and characteristic coupling length @xmath116 and then evolves under the hamiltonian with @xmath117 . we study the time - dependent two - point correlation functions and the time evolution of the energy contained in one cavity . since , for these expectation values , a microcanonical mixed state or a pure state of macroscopically well - defined energy are , in the thermodynamic limit , equivalent to a canonical ensemble @xcite , we evaluate canonical averages in the following . and @xmath9 is different in space - time regions @xmath118 , @xmath119 , , see the text . [ fig : figexp ] , scaledwidth=65.0% ] to obtain the field - field correlations at any time @xmath120 , we write the time - evolved field operators in terms of the creation operators @xmath121 corresponding to @xmath116 as @xmath122 \label{dxphi } \\ \pi ( x , t ) & = & p + i\sqrt{\frac{c}{2 } } \sum_{k > 0 } e^{-k\lambda/2 } \left [ \omega_{k}(x , t ) a_k^\dag - \omega^*_{k}(x , t ) a_k^{\phantom{\dag } } \right ] . \label{pi } \end{aligned}\ ] ] the coefficients @xmath123 and @xmath124 are given by @xmath125 \label{thetak } \\ \omega_{k}(x , t ) & = & k^{1/2 } \sum_{q>0 } ( k\|q ) \varphi_q(x ) \left [ \cos(cqt ) + i \frac{q}{k } \sin(cqt ) \right ] \label{omegak}\end{aligned}\ ] ] where @xmath126 , @xmath127 and @xmath128 are the eigenmodes corresponding to @xmath43 and @xmath116 , respectively . if @xmath128 is an even function of @xmath102 then only the even @xmath127 contribute to the sums and . moreover , in this case , since the even eigenmodes do not depend on the characteristic coupling length , these expressions simplify to @xmath129 and @xmath130 . for odd @xmath128 , it is instructive to first consider the two limiting cases of vanishing and large @xmath43 . for @xmath54 , the odd eigenmodes are @xmath131 with the wavenumbers @xmath56 where @xmath57 . by evaluating @xmath132 and using the relation , we obtain @xmath133 \nonumber \\ & ~&-\frac{a_k}{2 } k^{1/2 } e^{-ikl } \big [ ( -1)^{\lfloor \tau-{\bar x } \rfloor } e^{2ikl \epsilon(\tau-{\bar x } ) } - ( -1)^{\lfloor \tau+{\bar x } \rfloor } e^{2ikl \epsilon(\tau+{\bar x } ) } \big ] \label{omegakd0}\end{aligned}\ ] ] where @xmath34 runs over all integers , @xmath134 , @xmath135 , @xmath136 denotes the largest integer smaller than @xmath137 , and @xmath138 . inspection of this expression shows that the field @xmath3 is the superposition of two waves travelling with velocity @xmath1 and reflected with no phase shift at @xmath139 . for a given @xmath102 , @xmath140 takes two different values as time goes on , see fig [ fig : figexp ] . we find @xmath141 & \hbox { in regions a , a ' , \ldots } \nonumber \\ & = e^{i(\pi-2kl)(p+1/2 ) } \sin(kx ) & \hbox { in regions b , b ' , \ldots } \label{eomegak}\end{aligned}\ ] ] where @xmath142 in @xmath118/@xmath119 , @xmath8 in @xmath143/@xmath144 , . the space - time regions @xmath118 , @xmath119 , @xmath143 , @xmath144 , do not depend on the wavenumber @xmath145 . as we will see below , this independence plays an essential role in the behavior of physical properties such as the field - field correlations or the cavities energies . for the field @xmath9 , the coefficient is given by a similar expression . for large @xmath43 , we obtain @xmath146 . \label{omegakdinf}\ ] ] here , the field @xmath3 is the superposition of two waves reflected at @xmath58 and @xmath0 ( @xmath38 ) in the right ( left ) cavity . the expression becomes @xmath147 & \hbox { in regions a , a ' , \ldots } \nonumber \\ & = e^{-2ikl(p+1/2 ) } \mathrm{sgn } ( x ) \cos(kx ) & \hbox { in regions b , b ' , \ldots } \label{eomegakinf}\end{aligned}\ ] ] where @xmath142 in @xmath118/@xmath119 , @xmath8 in @xmath143/@xmath144 , . contrary to the above case , the field @xmath3 is always discontinuous at @xmath58 . the field @xmath9 satisfies @xmath148 , i.e. , the reflection is perfect at @xmath149 and @xmath46 . for a finite value of @xmath43 , the fields @xmath3 and @xmath9 are also superpositions of propagating waves but which are , contrary to the above studied special cases , both transmitted and reflected at @xmath58 . we write the coefficients @xmath124 and @xmath150 as @xmath151 \label{omegakgen } \\ \theta_k ( x , t ) & = & i \left ( \frac{d}{d_0 } -1 \right ) a_k \cos ( kl ) k^{-1/2 } \left [ \gamma_k^+(x , t ) - \gamma_k^-(x , t ) \right ] \label{thetakgen } \\ \gamma_k^\pm(x , t ) & = & \sum_q \frac{q a_q^2}{k - q } \frac{e^{iq(ct\pm \|x\| ) } } { dq\pm 2i } \label{gammadef}\end{aligned}\ ] ] where the sum runs over the positive and negative solutions of . to evaluate the functions @xmath152 , we first note that , since @xmath24 is a solution of , the summand in can be multiplied by a factor @xmath153^m$ ] where @xmath154 is any integer . we then rewrite @xmath152 as @xmath155 where @xmath156 and @xmath157 is a contour enclosing the real axis . expressing the above integral in terms of the residues of the integrand poles on the real axis leads to . for times @xmath120 between @xmath158 and @xmath159 where @xmath160 , the upper ( lower ) part of the contour @xmath157 can be closed in the upper ( lower ) half plane . moreover , for positive @xmath154 , the integrand in has poles only on the real axis and at @xmath161 . thus , for @xmath162 , the first term of the expression is determined by the residue of this last pole if @xmath154 is chosen as the largest integer smaller than @xmath163 . it can be shown that the functional dependence of this term on @xmath164 is @xmath165 where @xmath166 ( @xmath167 ) is a polynomial of order @xmath168 ( @xmath154 ) , see [ ab ] . the functions @xmath124 and @xmath150 change discontinuously at @xmath169 . for @xmath170 , they then reach the values given by the second term of in a time of order @xmath171 . for @xmath54 , one retrieves the dirac delta functions of and the second term of gives the expression . in the opposite limit , @xmath172 , the first term of is negligible with respect to the second one which leads to the expression . contrary to the limiting cases of vanishing and large @xmath43 studied above , since the successive reflections and transmissions at @xmath58 deform the propagating waves , @xmath173 and @xmath174 are not strictly periodic here . correlation functions @xmath175 ( full line ) and @xmath176 ( dashed line ) as functions of @xmath81 for @xmath177 , @xmath178 , @xmath179 , @xmath180 , and @xmath181 . the discontinuities correspond to @xmath182 , see fig . [ fig : figexp ] . the spatial correlations of the field @xmath111 are indistinguishable from that of the field @xmath3 ( full line ) . the dotted line is the thermal equilibrium correlation function .,scaledwidth=65.0% ] with the notations introduced in and , the two - point correlation functions of the cavities read @xmath183 since the functions @xmath150 and @xmath124 present discontinuities at positions independent of @xmath145 , as discussed above , the field - field correlations at a given time change abruptly at @xmath184 and @xmath185 where @xmath186 , @xmath187 , see fig . [ fig : corr ] and [ fig : corr2 ] . the results shown in figures [ fig : corr ] and [ fig : corr2 ] are obtained by numerically solving and evaluating the exact expressions - and - . for small @xmath43 , the first term of contributes significantly to the correlations - only if @xmath102 or @xmath103 is close to @xmath188 , as shown in fig [ fig : corr ] . this term becomes negligible in the large @xmath43 limit , see fig . [ fig : corr2 ] . thus , the field - field correlations are essentially described by the second term of for small and large @xmath43 . keeping only this term and using the properties @xmath189 , we obtain @xmath190 where the second sum runs over the positive and negative solutions of with the characteristic coupling length @xmath116 . the coefficients @xmath191 and @xmath192 are given by @xmath193 where @xmath194 and the regions @xmath118 , @xmath119 , @xmath143 , are presented in fig . [ fig : figexp ] . therefore , the approximate correlation function , for given @xmath102 and @xmath103 , is periodic with period @xmath195 . when @xmath102 and @xmath103 are both in a region of type @xmath118 , the expression is identical to the initial condition . as discussed after equation , the second and fourth terms of are negligible except for @xmath102 and @xmath103 near a cavity end , and hence is practically equal to for @xmath102 and @xmath103 both in a region of type @xmath119 , see fig . [ fig : corr ] . when @xmath102 and @xmath103 are in regions of different types , can differ considerably from . for example , for cavities perfectly coupled ( @xmath54 ) but initially uncoupled ( @xmath196 ) , @xmath197 in this case and thus the spatial correlations of the field @xmath3 are reduced by a factor of two with respect to the initial condition . it is interesting to note that even for uncoupled cavities ( @xmath89 ) , the sudden change of the characteristic coupling length causes a time - evolution of the correlation functions as @xmath198 is equal to @xmath8 only for @xmath196 . two - point correlation function of the field @xmath3 for @xmath199 , @xmath200 and @xmath201 . the other parameters are as in fig.[fig : corr ] . the curves for @xmath202 are indistinguishable from one another.,scaledwidth=65.0% ] for the other correlation functions , we find @xmath203 where @xmath191 and @xmath192 are given by , and , for @xmath204 , @xmath205 where the sum runs over the solutions of , the upper sign is for @xmath206 , and @xmath207 contrary to the correlations and of a field with itself , the correlations between the fields @xmath3 and @xmath9 given by vanish when @xmath102 and @xmath103 are in the same region , see fig . [ fig : corr ] . for @xmath102 and @xmath103 not too close to a cavity end , @xmath208 and @xmath209 where the sign depends on the relative positions of @xmath39 , @xmath102 and @xmath103 . we have seen in the previous section that there is no correlation between the two fields and that the correlations of a field with itself are essentially the same everywhere at equilibrium . switching the characteristic coupling length to another value induces correlations between the two fields and affects the whole system even far away from the interface between the two cavities . , @xmath178 and @xmath85 and @xmath84 . in the left part of the figure , @xmath210 for @xmath85 and @xmath211 for @xmath212 . the dotted lines correspond to the small @xmath43 approximation . [ fig : hct],scaledwidth=75.0% ] we study here the time evolution of the energy content of the cavities . for that purpose , we first note that the energy density @xmath213 is even with respect to @xmath102 . this symmetry property leads to @xmath214 the time dependence of the cavities energies is thus simply related to that of the coupling energy . as the eigenmodes of the total hamiltonian @xmath215 satisfy @xmath216/d$ ] , the time - evolved coupling hamiltonian is proportional to @xmath217 and hence can be expressed in terms of the functions . we obtain @xmath218 the even eigenmodes do not contribute to this sum since their derivatives vanish at @xmath58 . for the odd eigenmodes , we have seen above that , for @xmath170 , the first term of the expression contributes only for @xmath120 close to @xmath219 where @xmath154 is an integer . the second term of gives , using @xmath220 , @xmath221 where the sum runs over the wavenumbers @xmath145 corresponding to odd @xmath222 . for a small but finite @xmath43 , the coupling energy deviates from this value only for short periods of time when @xmath120 is much smaller than a characteristic time @xmath223 which increases with decreasing @xmath43 . in this short time regime , the energy density is essentially constant except in two regions wich propagate with velocity @xmath1 and are reflected and transmitted at @xmath58 , @xmath0 and @xmath38 . the multiple peak structure shown in left part of fig . [ fig : hct ] results from the preceding reflections and transmissions at @xmath58 . the temporal evolution of @xmath224 becomes more complicated for @xmath225 , see fig . [ fig : hct ] . for @xmath172 , the contributions of the first and second terms of to the functions @xmath226 are of the same order and hence the approximate expression does not apply . for @xmath89 , @xmath217 vanishes , as mentioned after . however , because of the factor @xmath43 in , some care must be taken in the evaluation of @xmath224 in this limit . for large @xmath43 , the wavenumbers @xmath24 are practically the multiples of @xmath48 except the lowest one @xmath227 . the terms of the sum are thus essentially equal to their infinite @xmath43 values except the first one and @xmath228 varies sinusoidally with a finite amplitude . as a function of frequency @xmath229 for the correlation function @xmath230 with @xmath231 , @xmath232 , @xmath233 , @xmath234 and @xmath235 . [ fig : spectre],scaledwidth=65.0% ] the time - dependent average energies @xmath236 , @xmath237 and @xmath224 , and the correlation functions - can be written as @xmath238 where @xmath239 stands for @xmath240 , , the sums run over the positive and negative @xmath24 corresponding to the characteristic coupling length @xmath43 , @xmath241 are appropriate coefficients , @xmath242 and @xmath243 . we remark that the even eigenmodes @xmath17 contribute only to the steady component of @xmath244 , see the discussion after equation . the expectation values of the single - particle observables of an isolated perfect gas , its particle number density at a given point for instance , are also given by expressions of the form with the single - particle energies @xmath245 @xcite . consider as an example a one - dimensional perfect gas confined in a box of length @xmath0 . in this case , the single - particle energies are @xmath246 where @xmath154 is the particle mass and @xmath34 an integer . the frequencies @xmath247 are thus regularly spaced and @xmath244 is periodic with period @xmath248 . however , in the joule expansion studied in ref . @xcite , for times @xmath249 , the function @xmath250 in the integral expression of the gas density profile , can be approximated as a continuous function plus a term @xmath251 and @xmath244 essentially relaxes from its initial value to @xmath252 . the situation is radically different for the free field system considered in this paper . there obviously also exists a @xmath253 contribution to @xmath254 but @xmath244 does not relax . first of all , we observe that , in the limiting cases of vanishing and large @xmath43 , the function @xmath250 is a sum of equally spaced dirac delta functions and @xmath244 is strictly periodic with period @xmath195 as shown by and . for a finite @xmath43 , the frequencies @xmath255 are not exactly equal to multiples of @xmath256 but most of them are close to these values , see fig . [ fig : spectre ] . as it is clear from equation , the spacing between consecutive frequencies @xmath257 varies and goes to zero for @xmath258 where @xmath259 . we see on fig . [ fig : spectre ] that @xmath260 in this limit . time regimes may then exist in which @xmath254 can be approximated as a sum of smooth functions with finite support . however , these functions clearly vanish at the right end of their support interval but not at the left one . consequently , their fourier transforms do not vanish fast enough at long times to lead to a relaxation behavior of @xmath244 . we find similar functions @xmath250 for the energy @xmath224 with the noticeable difference that the coefficients @xmath241 corresponding to the lowest frequencies @xmath261 are dominant in the large @xmath43 limit , giving rise to the sinusoidal behavior discussed at the end of the previous section . in this paper , we have studied how changing the thermal contact between two quantum systems whose elementary excitations are noninteracting bosons , affects these systems . more precisely , our model consists of two one - dimensional cavities of equal length coupled to each other by one of their ends . the coupling we have considered is thermodynamically small , as shown by the fact that the equilibrium total energy and correlation functions are practically independent of its strength , and allows to describe from perfectly coupled to completely uncoupled systems . we have seen that a sudden change of this thermal contact has a considerable impact on the correlation functions of the two cavities . they do not return to equilibrium but essentially oscillate with a period equal to twice the time required for a signal to propagate from one end of a cavity to the other . their exact time evolution is mainly controlled by the final coupling strength which determines the reflection and transmission coefficients at the interface between the two systems . we found a similar behavior for the energy content of the cavities . for weak coupling between them , it varies sinusoidally with a period set by the coupling strength . the absence of relaxation behavior is not simply a consequence of the fact that the considered hamiltonian is quadratic . the joule expansion of a perfect quantum gas has , for example , been obtained in ref . as discussed in the previous section , this difference in behavior comes from qualitative differences in the elementary excitation energy spectrum . clearly , the simple dispersion relation of our one - dimensional continuous field model , plays a crucial role in the temporal quasiperiodicity obtained . it would be thus interesting to study higher - dimensional and lattice free field systems . are interactions between the bosonic excitations necessary to ensure the stability of the thermal equilibrium ? to evaluate the energy for vanishing and large @xmath43 , we make use of @xmath262 where @xmath34 runs over all integers and @xmath263 is any function . with this relation , we obtain , for large @xmath43 , @xmath264 where @xmath265 using the expansion @xmath266 and , we find , in the limit @xmath267 , @xmath268^{-2}$ ] for @xmath269 , and @xmath270 . replacing these expressions into leads to with . for @xmath54 , the wavenumbers of the odd eigenmodes are @xmath56 whereas @xmath33 for the even ones . the contributions of the odd and even eigenmodes are then different . we obtain @xmath271 where @xmath272 , which gives . the polynomials @xmath273 and @xmath274 defined after are given by @xmath275^{-1}. \nonumber\end{aligned}\ ] ] the above integral is determined by the residue of the function @xmath263 at @xmath161 . we write @xmath276^{-1 } \nonumber \\ & = & \sum_{p , r \ge 0 } \frac{(i\tau)^p}{p ! } a_r \epsilon^{p+r - m } .\end{aligned}\ ] ] with these notations , becomes @xmath277 for @xmath278 ( @xmath279 ) . we find , for example , @xmath280 .\end{aligned}\ ] ]	in this paper , we address the issue of the stability of the thermal equilibrium of large quantum systems with respect to variations of the thermal contact between them . we study the schrdinger time evolution of a free bosonic field in two coupled one - dimensional cavities after a sudden change of the contact between the cavities . though the coupling we consider is thermodynamically small , modifying it has a considerable impact on the two - point correlation functions of the system . we find that they do not return to equilibrium but essentially oscillate with a period proportional to the length of the cavities . we compare this coupled cavities system with the perfect gas which is described by similar expressions but behaves very differently .
over a decade ago , slonczewski @xcite and berger @xcite predicted that an electrical current induces a torque on the magnetization of a ferromagnetic metal , and subsequent research has since identified distinct contributions called the reactive spin transfer torque and the dissipative spin transfer torque , sometimes referred to as adiabatic and non - adiabatic , respectively . @xcite progress on the understanding of this effect in metals has been steady , @xcite and the field has been stimulated by applications in spintronics and nanotechnology , as a way to manipulate magnetization and thus information . current - induced spin torques arise from a small mismatch between the spin polarization of conduction electrons and the magnetization present throughout a material , and reflect the nonlocal nature of magnetization dynamics in inhomogeneous systems . @xcite they are the converse of processes such as giant magnetoresistance ( gmr ) . the calculation of spin torques is equivalent to finding the conduction - electron spin density in an electric field , and in metals they can be easily expressed as the divergence of the spin current . @xcite recently spin torque - related effects have been investigated in metal - based systems and nanomagnets , e.g. , spin - torque driven ferromagnetic resonance , @xcite spin torques in nanomagnets @xcite and in continuously variable magnetizations , @xcite the @xmath0-@xmath1 interaction in inhomogeneous ferromagnets , @xcite current - induced magnetization dynamics , @xcite and domain wall motion . @xcite despite these efforts , spin torques have been little studied in materials displaying some of the richest physics and technological promise , namely ferromagnetic semiconductors . the most intensely studied ferromagnetic semiconductors are mn - doped iii - v compounds , which fall into two classes : zincblende lattices , such as ( ga , mn)as and wurtzite lattices , such as ( ga , mn)n . the former have been much investigated @xcite and will be the subject of this work , while the latter are awaiting further experimental developments . @xcite recent experiments have succeeded in fabricating ( ga , mn)as spin transfer torque devices , @xcite while domain wall motion @xcite and resistance @xcite have been addressed , and spin transfer physics was the subject of a recent review . @xcite ferromagnetic semiconductors differ profoundly from ferromagnetic metals . being @xmath2-type , the carriers are holes described by an effective spin-3/2 and subject to a strong spin - orbit interaction . the hole spin is thus not conserved . the band hamiltonian is the luttinger hamiltonian , @xcite which , in the spherical approximation , is composed of a series of multipoles in spin space . @xcite the hole spin is a dipole , while the spin - orbit interaction is a quadrupole , and the precession of the spin in a quadrupole field is highly nontrivial . @xcite magnetism is due to localized mn moments , with the exchange interaction mediated by itinerant holes . in the mean - field approximation this interaction is taken into account as an effective field ( a dipole ) acting on the carriers and produces a splitting comparable in magnitude to the fermi energy . @xcite the combined effect of this dipole exchange field and the quadrupole spin - orbit interaction is not a simple additive problem . @xcite these materials are often in the weak momentum scattering limit , @xcite i.e. , @xmath3 , where @xmath4 is the fermi energy and @xmath5 is a characteristic scattering time . due to the fast spin precession as a result of the spin - orbit interaction the relaxation time approximation is inappropriate to describe spin dynamics . the carrier spin is not conserved which introduces arbitrariness in the definition of the spin current . @xcite therefore , spin torques can not be identified , as in metallic systems , @xcite with the divergence of the spin current . finally electrically - induced spin densities @xcite in bulk nonmagnetic zincblende semiconductors are forbidden by symmetry , raising the question of what form the spin density may take in such a system when the symmetry is lowered by the magnetization . in this paper we present a general theory of current - induced spin torques in zincblende iii - v ferromagnetic semiconductors , which , to our knowledge , is the first of its kind . our study is based on the luttinger hamiltonian @xcite and it assumes @xmath6 , a time - independent magnetization with small spatial gradients , zero temperature , no compensation and short - ranged impurity potentials ( justified by the large carrier densities in ferromagnetic semiconductors and the short - range nature of the exchange interaction @xcite ) . we take the spin - orbit interaction to be the _ dominant _ term and treat the exchange field in first - order perturbation theory . this is a good approximation for the lower end of mn concentrations ( @xmath7 ) . we neglect also quantum interference effects such as weak localization . in recent years a number of microscopic theories of spin torques in ferromagnetic metals have been developed . @xcite a common strategy is to begin with a uniformly magnetized state and consider small perturbations around it . the method used in this work is slightly different , although the physics is the same and no difference is expected in the final results . we assume a magnetization that is a function of position and has small gradients , and we work to first order in the gradient of the magnetization . we would like to note that a recent study has been published which considers spin torques in metals starting from a spin continuity equation . @xcite the outline of this paper is as follows . the next section contains a derivation of the kinetic equation that will be used in the remainder of the paper . we begin from the quantum liouville equation and project it in momentum space , then introduce so - called wigner coordinates and derive the equation satisfied by the wigner distribution . in section iii this equation is solved in the presence of an electric field , and in section iv the spin density induced by the electric field is found , which gives the spin torque acting on the magnetization . the form and implications of the results are discussed , and their applicability to gamnas is demonstrated . finally , the domain wall velocity as a result of the spin torque is estimated in sec . [ sec : dws ] . the typical setup for a spin torque experiment consists of two slabs of ferromagnetic material with non - collinear magnetizations , separated by a tunnel barrier . since the magnetizations of the two slabs are non - collinear there is a region near the interface over which the magnetization changes . to determine the continuum limit of this setup one can begin by visualizing a large number of slabs put together , with slight variations in the direction of the magnetization of each slab . then one can imagine the interfaces between the slabs disappearing , leaving one large sample with an inhomogeneous magnetization , in such a way that the gradient of the magnetization varies little over distances comparable to the lattice spacing . the gradient expansion in the magnetization that follows from this procedure is valid as long as the length scale on which the magnetization varies is much longer than the relevant length scales of the carriers , i.e. , the fermi wavelength and mean free path . spin torques appear when an electrical current flows through such a material . in a ferromagnetic semiconductor the magnetization is a result of the mn ions which interact by means of the exchange coupling mediated by itinerant holes . the holes themselves have a spin polarization , and in equilibrium the hole spin polarization follows the magnetization . when an electrical current flows through the sample one can think for example of a hole which is taken from position @xmath8 , where its spin is parallel to the local magnetization at @xmath8 , and transporting it to @xmath9 . the magnetization at @xmath10 is slightly different from the magnetization at @xmath8 , so a torque is exerted by the itinerant hole on the magnetization . what makes the situation in ferromagnetic semiconductors more difficult and more interesting is that while the hole is moving from @xmath8 to @xmath10 it is subject to the strong spin - orbit interaction , which acts to randomize its spin . in order to determine the effect of spin - orbit interactions , which are wave vector - dependent , on the itinerant holes and ultimately on the inhomogeneous magnetization we need to study the kinetic equation , which takes into account both the momentum dependence and the position dependence . in this section we will derive a kinetic equation suitable for describing inhomogeneous ferromagnetic semiconductors in an electric field . we consider the system to be described by a density operator @xmath11 , which obeys the quantum liouville equation @xmath12 = 0.\ ] ] the total hamiltonian @xmath13 contains contributions due to the band hamiltonian @xmath14 , the scalar impurity potential , the exchange interaction between delocalized holes and localized mn moments , and the electric field . these will be given below . the liouville equation is projected onto a set of states @xmath15 of definite wave vector @xmath16 and spin @xmath0 , which are assumed to be bloch functions and eigenstates of the luttinger hamiltonian @xmath14 . the matrix elements of @xmath17 in this basis are @xmath18 and are treated as matrices in spin space . @xmath14 is diagonal in @xmath16 and its matrix elements in this basis are @xmath19 @xcite @xmath20,\ ] ] where @xmath21 is a vector of spin-3/2 matrices . the term proportional to the luttinger parameter @xmath22 gives the hole kinetic energy . for @xmath23 , the term proportional to @xmath24 separates the heavy hole ( hh ) and light hole ( lh ) states , i.e. , it is the spin - orbit coupling that plays a central role in the present analysis . the mn@xmath25 ions give rise both to a net magnetic moment , through the hole - mediated exchange interaction , and to scattering , which has a scalar part and a spin - dependent part . these are contained in the hamiltonian @xmath26 , @xmath27,\ ] ] where the sum runs over the positions @xmath28 of the mn@xmath25 ions , with @xmath29 the mn spin . we approximate the interactions represented by @xmath26 as short - ranged , so that @xmath30 and @xmath31 , with @xmath32 the exchange constant between the localized mn moments and the itinerant holes , and @xmath33 the sample volume . the matrix elements of @xmath26 in the basis @xmath34 are decomposed into a part @xmath35 diagonal in @xmath16 , which gives the net magnetization @xmath36 , @xmath37 where @xmath38 is the number of mn@xmath25 ions and we assume all mn spins polarized setting @xmath39 , and a part off - diagonal in @xmath16 , which causes spin - dependent scattering and will be given below . we will concentrate in this work on the case in which the exchange splitting is smaller than the spin - orbit coupling at the fermi energy , i.e. , @xmath40 , where @xmath41 is the fermi wave vector , and we work to first order in @xmath42 . in studying inhomogeneous magnetization one must account for real - space as well as wave - vector dependence , which is accomplished by defining a wigner distribution . the wigner function corresponding to the one - particle density matrix @xmath43 is @xmath44 where @xmath45 and @xmath46 . the next step is to derive an equation describing the time evolution of the wigner distribution @xmath47 . the kinetic equation for the wigner function @xmath48 is obtained by projecting the quantum liouville equation onto the basis states @xmath49 , then making the transformation ( [ eq : wigin ] ) . the hamiltonian @xmath13 is diagonal in wave vector . the first step , in which no approximations have been made , gives us the liouville equation in terms of the so - called wigner coordinates @xmath50 and @xmath51 @xmath52 to obtain a transparent kinetic equation it is necessary to expand @xmath53 around the wave vector @xmath50 , which requires some care . in this article we are working in a basis in which the functions depend on the wave vector @xmath50 , and we require a formulation of the kinetic equation that is manifestly gauge covariant . this means that the kinetic equation should not acquire additional terms if the basis functions are subjected to a @xmath54-dependent rotation . the end result of this requirement is that ordinary @xmath50-derivatives are replaced by covariant derivatives ( @xmath8-derivatives , denoted by @xmath55 , remain unchanged since the basis does not have position dependence . ) the covariant @xmath54-derivative is defined by @xmath56 $ ] , where the gauge connection matrix @xmath57 . all this implies that in our derivation we must replace the ordinary derivatives by covariant derivatives . expanding @xmath58 equation ( [ frho ] ) can be written as @xmath59 + \frac{1}{2\hbar}\ , \ { \frac{dh_{{\bm q}}}{d{\bm q } } \cdot \nabla f_{\bm q } \ } = - \frac{i}{\hbar } \sum_{{\bm \kappa } } \int \frac{d^3q}{(2\pi)^3}\ , e^{i{\bf q}\cdot{\bf r } } \big(u_{{\bm q}_+{\bm \kappa } } \rho_{{\bm \kappa}{\bm q}_- } - \rho_{{\bm q}_+{\bm \kappa } } u_{{\bm \kappa}{\bm q}_-}\big ) , \end{array}\ ] ] where @xmath60 denotes an anti - symmetrized dot product , that is the symmetrized scalar product between vectors @xmath61 and @xmath62 is given by @xmath63 . the scattering term represented by the rhs is dealt with in the appendix . in a constant and uniform electric field the total hamiltonian contains an additional term containing the electromagnetic potential @xmath64 , where @xmath65 is the position operator . this term is diagonal in real space . following the spirit of the derivation presented above , this term appears on the right side of the kinetic equation in the same way as @xmath66 , and is expanded as follows @xmath67 \| { \bm q}_- \rangle } \approx -\frac{1}{\hbar}\ , { \bm \nabla } v \cdot \frac{df_{{\bm q}}}{d{\bm q}}.\ ] ] the spatial gradient of the external electrical potential is equal to the electric field @xmath68 . in this work we will be studying the response of the system to linear order in the electric field . when formulating a kinetic equation , which takes into account the variation of the wigner function in real space as well as in momentum space , it is necessary to single out the length and wave - vector scales relevant to the problem under study . in the work at hand we consider carriers which are delocalized in real space and are described by bloch states , for which the wave vector is a good quantum number . nevertheless it must be borne in mind that the carrier occupies a finite range of real and momentum space , denoted by @xmath69 and @xmath70 respectively , which are determined in such a way as to be consistent with the heisenberg uncertainty principle . in the course of a scattering event in which a carrier with wave vector @xmath50 interacts with the potential of an impurity and its wave vector changes from @xmath50 to @xmath71 , it is necessary as well as physical to assume that the wave - vector spread @xmath70 associated with the carrier size is m ! uch smaller than the typical momentum transfer in scattering processes @xmath72 . furthermore , it is assumed that the magnetization @xmath73 varies over length scales much larger than interatomic separations . with these assumptions , the kinetic equation in an electric field @xmath74 takes the form ( in agreement with the form found by carruthers and zachariasen @xcite ) @xmath75 + \frac{1}{2\hbar } \left\ { \frac{d}{d{\bm q}}\ , ( h_v + h_{pd } ) \cdot \nabla f \right\ } - \frac{1}{2\hbar } \left\ { \nabla h_{pd } \cdot \frac{df}{d{\bm q } } \right\ } + \hat j ( f ) = \sigma_e,\ ] ] where @xmath76 is the covariant form of the usual source term due to @xmath74 . the term @xmath77 represents the scattering term , which is discussed in detail in the appendix . the scattering term takes into account the effect of the potential @xmath78 , which represents the part of the hamiltonian @xmath79 which is off - diagonal in wave vector . an explicit form for the scattering term will be given below when we discuss the solution of the kinetic equation in an electric field . the equilibrium distribution @xmath80 is the solution to eq.([eq : equikineq ] ) in the absence of external fields , @xmath81 . to leading order in @xmath82 this solution is @xmath83 , with @xmath84 the fermi - dirac function . it is straightforward to check that this form of the wigner function satisfies the kinetic equation ( [ eq : equikineq ] ) when the rhs is equal to zero . the form of his solution shows that in equilibrium the spin polarization of the holes follows the magnetization of the mn , which is contained in the exchange part of the hamiltonian @xmath35 . next , in the linear response regime , we search for a solution of the kinetic equation for nonzero @xmath85 , which will yield the spin density induced by @xmath74 . since the spin density induced by the electric field will be a function of position and will in general not be parallel to the local magnetization , this will immediately give the spin torque exerted by the conduction holes on the magnetization . the method we use to solve the kinetic equation is as follows . first , we divide every matrix @xmath86 in the problem into @xmath87 , where @xmath88 has elements only within the hh and lh subspaces , while @xmath89 has matrix elements only between these subspaces . schematically this can be summarized by @xmath90 one compelling advantage of this decomposition is that commutators and anticommutators of matrices belonging to either the _ in _ or _ out _ sectors do not mix these sectors . the following list covers all the possible combinations of commutators and anticommutators of matrices belonging to either the @xmath91 or the @xmath92 sectors @xmath93 = & \displaystyle out \\ [ 3ex ] \displaystyle [ in , in ] = & \displaystyle in \\ [ 3ex ] \displaystyle [ out , out ] = & \displaystyle out \\ [ 3ex ] \displaystyle \ { in , out \ } = & \displaystyle out \\ [ 3ex ] \displaystyle \ { in , in \ } = & \displaystyle in \\ [ 3ex ] \displaystyle \ { out , out \ } = & \displaystyle in . \end{array}\ ] ] another advantage of this decomposition is that it aids us in constructing a physical picture of spin torques and their relation to spin precession . the decomposition into an @xmath91 and an @xmath92 sector in effect singles out spin precession as a result of the spin - orbit interaction . the _ in _ sector of the density matrix represents spins that are stationary under the action of the spin - orbit interaction , or alternatively the fraction of the spins that are in eigenstates of @xmath94 . the _ out _ sector on the other hand represents spins that precess under the action of the spin - orbit interaction . this decomposition determines which spin torques are due to the hole spin precession , which , unlike the precession of spin-1/2 electrons , can not be attributed to an effective magnetic field . @xcite being in the weak momentum scattering regime @xmath95 , we do not consider scattering in the _ out _ sector or between the _ in _ and _ out _ sectors ( it can be shown that both of these terms yield corrections linear in @xmath82 ) . the wigner function @xmath96 has two parts , @xmath97 in the @xmath91 sector and @xmath98 in the @xmath92 sector , and the kinetic equation is broken down into two coupled equations for @xmath97 and @xmath98 [ eq : inout ] @xmath99 + \hat j(f^{in } ) & = & \sigma_e^{in } + \sigma_\mathrm{gr}^{in } \\ { \frac{\partial f^{out}}{\partial t } } + \frac{i}{\hbar}\ , [ h^{in } , f^{out } ] & = & \sigma_e^{out } + \sigma_\mathrm{gr}^{out}. \hspace{1em } \end{aligned}\ ] ] there are two source terms in each equation , namely @xmath100 and @xmath101 in the @xmath91 sector , and @xmath102 and @xmath103 in the out sector . to obtain these source terms one needs to expand all quantities in the gradient of the magnetization and keep terms to zeroth and first order in this gradient . to zeroth order in the gradient of the magnetization the source terms are @xmath104 , which are found simply by taking @xmath85 defined above and substituting @xmath105 for the wigner function . when the expansion is continued to the next order , the source terms linear in the gradient ( gr ) of the magnetization are [ eq : source ] @xmath106 . \end{aligned}\ ] ] to obtain eq . ( [ eq : source ] ) we have assumed a small spatial gradient @xmath107 implying a small variation @xmath108 , and we worked , as stated , to first order in @xmath109 . after some simplification we obtain for the scattering term acting on @xmath97 @xmath110,\ ] ] where the bar is an average over directions in momentum space , @xmath111 , @xmath112 and @xmath113 , @xmath114 and @xmath115 are the polar and azimuthal angles of @xmath50 ( analogously for @xmath116 ) , and @xmath117 is the carrier effective mass , which is @xmath118 in the hh subspace , and @xmath119 in the lh subspace . for simplicity and without loss of generality we choose @xmath120 and @xmath121 so that @xmath122 . we solve eqs.([eq : inout ] ) as follows : the equation for @xmath98 is first solved with @xmath102 as the initial source , and the solution @xmath123 thus obtained is substituted into @xmath124 and @xmath125 . the equation for @xmath97 is solved in an analogous fashion . the solutions to the equations for @xmath97 and @xmath98 involve expressions of the form @xmath126 , and @xmath98 is easily found . this is because in the _ out _ sector the product @xmath127 contains only functions of time of the form @xmath128 and @xmath129 , with @xmath130 the energy difference between the hh and lh bands when the magnetization is zero . the steady - state solution for @xmath98 therefore involves only a straightforward time integral of the kind customarily encountered in linear - response theories . the equation for @xmath97 takes more effort due to the presence of the scattering term and we only summarize the method here ( it is described in detail in a recent publication by two of us@xcite ) . the @xmath91 sector represents the part of the wigner function that is stationary under the action of @xmath94 . nevertheless , the full hamiltonian is @xmath131 , and the commutator @xmath132 $ ] is not zero . in a manner similar to the decomposition of @xmath96 into @xmath97 and @xmath98 , @xmath97 itself is split into a part that commutes with @xmath133 , and a part that does not . it can be shown @xcite that the commuting part yields a correction to the wigner function that is linear in @xmath5 while the non - commuting part gives a correction that does not depend on @xmath5 . however , we find that all contributions to @xmath96 average to zero over directions in momentum space except @xmath134 . this implies that all contributions from @xmath97 average to zero over directions in momentum space . @xmath134 gives rise to a spin density @xmath135 that is independent of scattering . it is discussed in detail below . the only contribution to the spin density in an electric field comes from @xmath134 . the three components of the spin density @xmath136 that this correction to the wigner function yields are [ eq : torque ] @xmath137 these equations are the central result of our work . the dimensionless quantities @xmath138 and @xmath139 , with @xmath140 are functions of the luttinger parameters @xmath22 and @xmath24 . for gamnas we find ( all @xmath141 ) @xmath142 , @xmath143 , @xmath144 and @xmath145 . the steady - state spin density is not collinear with the magnetization , so there will be a torque on the magnetization giving a precession frequency of magnitude @xmath146 . taking @xmath147 @xmath148 , @xmath149 kv / m , and estimating the change in the magnetization as 20@xmath150 over one lattice spacing , the time scale of this precession is 200 ns less than in metals , but @xmath73 itself is also typically one order of magnitude smaller . the fact that the spin torque comes only from @xmath134 implies that the steady - state spin density is due to precession under the action of both the spin - orbit interaction and the exchange field . the fraction of the spins that is conserved , which would yield a term @xmath151 , gives a contribution that averages to zero in momentum space . the quantities @xmath138 and @xmath139 decrease with increasing spin - orbit interaction ( given by @xmath24 ) , suggesting the spin - orbit interaction reduces the spin torque . this agrees with the finding that there is no electrically - induced spin density in the corresponding nonmagnetic systems , @xcite i.e. in the limit of large spin - orbit interaction @xmath152 . this limit is equivalent to restoring the spherical symmetry of the luttinger hamiltonian of eq.([eq : hlutt ] ) , which in ferromagnetic semiconductors is broken by the magnetization . an important difference from ferromagnetic metals is that , in eq.([eq : torque ] ) , there is no contribution from scattering , either scalar or spin - dependent . this fact indicates that the dominant spin torque in ferromagnetic semiconductors in the weak momentum scattering limit is intrinsic . this observation agrees with the results of jungwirth _ et al_. , @xcite who studied the anomalous hall effect in ferromagnetic semiconductors in the regime @xmath3 and found similarly that the role of scattering is secondary . it is also related to the absence of electrically - induced spin polarization in bulk nonmagnetic zincblende materials . generally , such a spin polarization is due to the fraction of spins that is conserved @xcite and is linear in @xmath5 , but this spin polarization is forbidden by symmetry in zincblende lattices . @xcite the magnetization breaks the cubic symmetry of the lattice and gives a steady - state spin density , but the term linear in @xmath5 still averages to zero . we come back to the comparison of our result with result found for ferromagnetic metals in the next section . we find that an electric field @xmath153 corresponds to the permutation @xmath154 in eq.([eq : torque ] ) . yet for a given orientation of @xmath74 , unlike in ferromagnetic metals , in ferromagnetic semiconductors there is no symmetry between the different components of the spin density for the following reason . in metals spin is conserved and spin torques can be derived phenomenologically directly from the landau - lifshitz - gilbert equation ( the so - called _ book - keeping _ argument @xcite ) . one assumes an itinerant spin passes a localized moment at @xmath8 , lines up with it , then moves on to another moment at @xmath10 and exerts a torque on this moment . this relates @xmath155 to @xmath156 and gives a simple vector - product form for @xmath157 . @xcite in ferromagnetic semiconductors the spin - orbit interaction acts to randomize the itinerant spin moving between @xmath8 and @xmath10 , and there is no simple relationship between @xmath158 and @xmath159 . such a book - keeping argument is thus not valid and there is no symmetry in the final expression for the spin density . we would like to comment on one last aspect of the relationship between the hole spin polarization and the magnetization in ferromagnetic semiconductors . the calculation presented in this work relies on a mean - field description of the magnetization and hole spin polarization . in this picture the itinerant holes are subject to an average magnetic field due to the mn@xmath25 ions , and the mn@xmath25 ions in turn are subject to the itinerant hole spin polarization , which can also be regarded as an average magnetic field.@xcite since it is assumed that the spin - orbit interaction has spherical symmetry , there is no easy axis for the magnetization in the absence of an electric field . however , once the electric field is applied it is natural to ask whether the direction of the electric field provides an easy axis for the magnetization , in other words whether the magnetization in the direction of the electric field increases . we find that this indeed is true , but the increase in the magnetization is second order in the ratio @xmath160 and is not significant . in ferromagnetic metals , in which spin - orbit coupling is negligible , angular momentum is conserved . as a result spin torques in these materials can be encapsulated into a set of simple , compact , rotationally - invariant vectorial expressions . in ferromagnetic semiconductors , in which spin - orbit interactions are usually the dominant energy scale , angular momentum is not conserved and the final expressions for the spin torques can not be expected to have rotational invariance . in principle spin - orbit interactions , which couple the spin and the lattice , should give magnetic anisotropy and anistropic spin torques as well . the anisotropy in our result for the spin density is thus a direct result of the intrinsic spin - orbit interactions . we shall assume a doping density @xmath161 , corresponding to @xmath162 , @xmath163 mev nm@xmath164 as discussed in ref . [ ] and the lattice constant @xmath165 . the fermi energy is found as @xmath166 \displaystyle \varepsilon_f = 1.633 \ , \frac{\hbar^2}{2m_0}\ , ( 3\pi^2n)^{2/3 } = 2.1 \times 10^{-20}\ , { \rm j } \end{array}\ ] ] and the heavy and light hole fermi wave vectors are @xmath167 and @xmath168 . the heavy hole and light hole masses are @xmath169 kg and @xmath170 kg . these numbers also give the magnitude of the effective field @xmath171 j , meaning that the ratio @xmath172 , so it is safe to do perturbation theory . we also want to work out @xmath173 . the fermi energy is @xmath174j , which means @xmath175 for any momentum scattering time @xmath176s . for example for @xmath177 we require @xmath178s , corresponding to a light - hole mobility of approximately @xmath179@xmath180/vs and a heavy - hole mobility of approximately @xmath181@xmath180/vs . thus the theory is on very firm ground even for extremely low mobilities . as an application of the central result in eq . ( [ eq : torque ] ) we calculate the spin torque on a domain wall , and the resulting domain - wall velocity . we choose the current and variation of magnetization in the @xmath182-direction . furthermore , we use @xmath183 such that eq . ( [ eq : torque ] ) reduces to @xmath184,\ ] ] with @xmath185 a unit vector in the @xmath182-direction . the spin - transfer torque that acts on the magnetization is given by @xmath186 using eq . ( [ eq : torque ] ) we rewrite this as an equation for a unit vector @xmath187 in the direction of magnetization , i.e. , @xmath188 , with @xmath189 the spin of one mn atom . we find that @xmath190,\ ] ] with the velocity @xmath191 given by @xmath192 the result for the current - induced torques in eq . ( [ eq : domwall : unitvector ] ) has the form of an anisotropic dissipative spin transfer torque . @xcite the reactive spin transfer torque contribution is equal to zero . these results are understood by noting that we have considered strong spin - orbit interactions , and that have done perturbation theory in the magnetization . it is common to define a dissipative coefficient @xcite @xmath193 such that @xmath194 , with @xmath195 the current density . because our result for the spin transfer torque is independent of @xmath5 , and because @xmath196 , we would find that @xmath197 , i.e. , resistivity - like . this is somewhat surprising as recent studies @xcite indicate that the gilbert damping constant @xmath198 , which is believed to be similar though not exactly equal to @xmath193 , predominantly has intra - band contributions that are conductivity - like . however , a direct comparison is not possible because in the present paper we perform an expansion in the magnitude of the magnetization whereas ref . [ ] calculates @xmath198 by determining the transverse response function . the velocity @xmath191 divided by the gilbert damping constant provides an estimate for the domain wall velocity @xmath199 , @xcite so that @xmath200 although sinova _ _ do not explicitly consider the regime of parameters quoted in the previous section , their calculations @xcite ( see also ref . [ ] ) suggest that the gilbert damping is very small @xmath201 in this regime . using this result we find that @xmath202 m / s , in agreement with experimental results for the domain - wall velocity . @xcite to investigate more quantitatively the effect of the anisotropy in the spin transfer torque , determined by the ratio @xmath203 , we consider specific model for a magnetic domain wall . we consider a thin film , in which there is a constant hard - axis anisotropy @xmath204 perpendicular to the film and an easy - axis anisotropy @xmath205 . within the model for a rigid domain wall proposed by tatara and kohno @xcite ( see also ref . [ ] ) , the domain wall is described by two collective coordinates : the position @xmath206 and the chirality @xmath207 . the chirality is the angle with which the magnetic moment in the center of the domain wall tilts out of the easy plane . using the results from ref . [ ] and ref . [ ] we find the equations of motion for the domain - wall collective coordinates . they are given by [ eq : domwall : eom ] @xmath208 where @xmath209 is the width of the domain wall . note that @xmath210 goes to zero for @xmath211 . note that , in addition usual dissipative spin transfer torque contribution to these equations discussed in earlier work , @xcite we find a chirality - dependent anisotropic contribution proportional to @xmath210 . the above equation can be solved analytically . from this we obtain the average drift velocity as a function of the applied electric field , as shown in fig . [ fig : driftvelocity ] . ( solid line ) and @xmath212 ( dashed line).,width=302 ] from this figure , we observe a walker - breakdown - like behavior , @xcite i.e. , the domain wall velocity reaches a maximum and then becomes smaller . physically , the breakdown is due to the transition of rigid motion to precessional motion of the domain wall and is well - known from field - driven domain - wall motion . our results are understood from the fact that dissipative spin - transfer torque enters the equations of motion for the domain wall in the same way as an external magnetic field . note that the anisotropy @xmath210 alters the result for the domain - wall velocity somewhat with respect to the isotropic ( @xmath212 ) situation , but plays no qualitatively important role . as a final remark , we note that in the calculations presented here we have neglected the effects of finite temperature @xcite and pinning of the domain wall . this , in addition to the fact that the experiments of yamanouchi _ @xcite are in a different regime of doping than considered here , makes a direct quantitative comparison not possible . in conclusion , we have established a microscopic theory of spin transfer in iii - v ferromagnetic semiconductors for the case of strong spin - orbit coupling . we have applied our results to the case of current - driven domain wall motion and have estimated the resulting domain - wall velocities . we find domain - wall velocities that are of the same order of magnitude as experiments , although the available experimental results @xcite are in a different regime of parameters than considered in this paper . therefore , a more quantitative comparison between theory and experiment is at present not feasible . we wish to acknowledge enlightening discussions with a. h.macdonald , m. tsoi , g. e. w. bauer , l. w. molenkamp , g. schmidt , b. j. van wees , m. van veenendaal , d. j. keavney and zhuge liang . the research at argonne national laboratory was supported by the us department of energy , office of science , office of basic energy sciences , contract no . de - ac02 - 06ch11357 . the research at utrecht university was supported by the netherlands organization for scientific research ( nwo ) and by the european research council ( erc ) under the seventh framework program ( fp7 ) . 10 j. slonczewski , j. mag . mag . mat . * 159 * , l1 ( 1996 ) . l. berger , phys . b * 54 * , 9353 ( 1996 ) . d. c. ralph and m. d. stiles , arxiv:0711.4608 ( 2007 ) ; p. m. haney _ et al . _ , arxiv:0709.3862 ( 2007 ) ; y. tserkovnyak _ et al . _ , arxiv:0709.2937 ( 2007 ) . ya . b. bazaliy _ et al . _ , phys . b * 57 * , 3213 ( 1998 ) . m. tsoi _ et al . _ , phys . lett . * 80 * , 4281 ( 1998 ) . m. d. stiles and a. zangwill , phys . b * 66 * , 014407 ( 2002 ) . j .- ph . ansermet _ et al . _ , ieee trans . mag . * 40 * , 358 ( 2004 ) . z. li and s. zhang , phys . lett . * 92 * , 207203 ( 2004 ) . a. thiaville _ et al . _ , europhys . lett . * 69 * , 990 ( 2005 ) . s. e. barnes and s. maekawa , phys . 95 * , 107204 ( 2005 ) . r. a. duine _ et al . _ , phys . b * 75 * , 214420 ( 2007 ) . f. piechon and a. thiaville , phys . b * 75 * , 174414 ( 2007 ) . h. kohno _ et al . _ , j. phys . jpn * 75 * , 113706 ( 2006 ) . y. tserkovnyak _ et al . _ , phys . b * 74 * , 144405 ( 2006 ) . y. tserkovnyak _ et al . _ , rev . phys . * 77 * , 1375 ( 2005 ) . j. n. kupferschmidt _ et al . _ , phys . b * 74 * , 134416 ( 2006 ) . x. waintal and o. parcollet , phys . . lett . * 94 * , 247206 ( 2005 ) . j. xiao _ et al . _ , phys . rev . b * 73 * , 054428 ( 2006 ) . a. rebei _ et al . _ , phys . b * 72 * , 064408 ( 2005 ) . j. fernandez - rossier _ et al . _ , b * 69 * , 174412 ( 2004 ) . g. tatara _ et al . _ , j. phys . . jpn . * 75 * , 064708 ( 2006 ) . t. jungwirth _ et al . _ , rev . phys . * 78 * , 809 ( 2006 ) . j. knig _ et al . _ in _ electronic structure and magnetism of complex materials _ , springer series in material sciences * 54 * , ( springer , 2003 ) ; j. knig _ et al . _ in _ interacting electrons in nanostructures _ , lecture notes in physics * 579 * ( springer , 2001 ) . d. j. keavney _ et al . _ , phys . lett . * 95 * , 257201 ( 2005 ) . j. wunderlich _ et al . _ , arxiv:0707.3329 ( 2007 ) . a. k. nguyen _ et al . _ , phys . rev . lett . * 98 * , 146602 ( 2007 ) . m. yamanouchi , d. chiba , f. matsukura , t. dietl , and h. ohno , phys . lett . * 96 * , 096601 ( 2006 ) . a. k. nguyen _ et al . _ , phys . rev . lett . * 97 * , 136603 ( 2006 ) . h. ohno and t. dietl , arxiv:0712.3247 ( 2007 ) . j. m. luttinger , phys . rev . * 102 * , 1030 ( 1956 ) . n. o. lipari and a. baldereschi , phys . rev . * 25 * , 1660 ( 1970 ) . d. culcer _ et al . _ , phys . * 97 * , 106601 ( 2006 ) . j. shi _ et al . _ , phys . lett . * 96 * , 076604 ( 2006 ) . e. l. ivchenko and g. e. pikus , jetp lett . * 27 * , 604 ( 1978 ) . g. tatara and p. entel , phys . b * 78 * , 064429 ( 2008 ) . h. ohno and t. dietl , j. magn . magn . mat . * 320 * 1293 ( 2008 ) . p. carruthers and f. zachariasen , rev . phys . * 55 * , 245 ( 1983 ) ; v. p. sillin , sov . jetp * 6 * , 945 ( 1957 ) . d. culcer and r. winkler , phys . b * 76 * , 245322 ( 2007 ) . t. jungwirth _ et al . _ , phys . rev . lett . * 88 * , 207208 ( 2002 ) . jairo sinove , _ et al . _ , , 085209 ( 2004 ) . ion garate and allan macdonald , arxiv:0808.3923v1 [ cond-mat.mtrl-sci ] . g. tatara and h. kohno , phys . 92 * , 086601 ( 2004 ) ; * 96 * , 189702 ( 2006 ) . duine , a.s . nez , and a.h . macdonald , phys . . lett . * 98 * , 056605 ( 2007 ) . schryer and l.r . walker , j. appl . * 45 * , 5406 ( 1974 ) . the scattering term @xmath213 is @xmath214\ , e^ { i \hat h t ' } \right]_{{\bm q}{\bm q}},\ ] ] with @xmath215 a regularization factor and the impurity average @xmath216 , where @xmath217 is the _ number _ of mn impurities and @xmath218 . the derivation of this general form of the scattering term is discussed in a recent paper @xcite and the notation will be explained in detail below . in terms of the wigner distribution the scattering term can be expressed as @xmath219 & \displaystyle - \frac{n_{mn}}{\hbar^2 } \int \frac{d^3\kappa}{(2\pi)^3 } \ , \int^{\infty}_0 dt'\ , e^{- \eta t'}\ , e^{- i h_{{\bm q } } t ' } \big ( u_{{\bm q}{\bm \kappa } } f_{\bm \kappa } - f_{\bm q } u_{{\bm q}{\bm \kappa } } \big ) \ , e^{i h_{{\bm \kappa } } t ' } u_{{\bm \kappa}{\bm q } } . \end{array}\ ] ] we must note that , in the approximation we are using , the scattering term acts only on @xmath97 , which brings about some simplifications . these become apparent if we look at the explicit form of this term and note that , because it involves only @xmath97 , this term commutes with the time evolution operators . @xmath220 - & \displaystyle \frac{n_{mn}}{\hbar^2 } \int \frac{d^3\kappa}{(2\pi)^3 } \ , \int^{\infty}_0 dt'\ , e^{- \eta t'}\ , \big ( e^{- i h_{{\bm q } } t'}\ , u \ , e^{i h_{{\bm \kappa } } t ' } f^{in}_{\bm \kappa } - f^{in}_{\bm q } e^{- i h_{{\bm q } } t'}u \ , e^{i h_{{\bm \kappa } } t ' } \big ) u . \end{array}\ ] ] in the approximation used in this paper , the hamiltonian entering the scattering term is the projected @xmath221 hamiltonian for each subspace . the scattering potential has two parts , one a scalar and one which is spin - dependent . taking into account also the exchange splitting of the bands , there are three contributions to the scattering term : scalar potential + kinetic energy ( @xmath222 ) , spin - dependent potential + kinetic energy ( @xmath223 ) , scalar potential + exchange energy ( @xmath224 ) . the former two sum up to @xmath225 = & \displaystyle \frac{n_{mn } m^q}{4\pi^2\hbar^3 } \int d\omega ' \ , \big ( \frac{1}{2}\ { u^2 , f^{in}_{\bm q } \ } - u f^{in'}_{\bm q } u \big ) = \frac{n_{mn } m^q}{\pi \hbar^3 } \ , \big ( \frac{1}{2}\ { \overline{u^2 } , f^{in}_{\bm q } \ } - \overline{uf^{in}_{\bm q}u } \big ) \\ [ 3ex ] = & \displaystyle \frac{1}{\tau}\ , \big ( \frac{1}{2}\ { \overline{\gamma^2 } , f^{in}_{\bm q } \ } - \overline{\gamma f^{in}_{\bm q } \gamma } \big ) \\ [ 3ex ] \displaystyle \frac{1}{\tau } = & \displaystyle \frac{n_{mn } \\ [ 3ex ] \displaystyle \overline{f^{in}_{\bm q } } = & \displaystyle \frac{1}{4\pi } \int d\omega ' \ , f^{in'}_{\bm q}. \end{array}\ ] ] we have used the notation @xmath226 . @xmath82 is a scalar and @xmath227 is a dimensionless matrix , which is written as @xmath228 , with @xmath229 . notice that @xmath230 has angular dependence because we are in the basis of eigenstates of the luttinger hamiltonian . thus these two contributions to the scattering term can be rewritten as @xmath231 = \frac{1}{\tau}\ , \big [ f + \ { \overline{(\gamma_s + \frac{\gamma_s^2}{2 } ) } , f \ } - ( \overline{f } + \overline{\ { \gamma_s , f \ } } + \overline{\gamma_s f \gamma_s } ) \big ] \\ [ 3ex ] = & \displaystyle \frac{f - \overline{f}}{\tau } + \frac{1}{\tau } \ , \ { \overline{\gamma_s + \frac{\gamma_s^2}{2 } } , f \ } - \frac{1}{\tau}\ , \overline { \ { \gamma_s , f \ } } - \frac{1}{\tau}\ , \overline{\gamma_s \ , f \ , \gamma_s}. \end{array}\ ] ] we think of @xmath5 as a characteristic scattering time . the explicit form of the potential , determined previously , is @xmath232 , \end{array}\ ] ] where @xmath233 . everything must be averaged over the impurity configuration as well as directions in momentum space , and then it needs to be transformed into the eigenstate basis and projected onto lh and hh . when we do that , the term linear in @xmath234 above contains only @xmath235 , which , when projected onto lh and hh gives something that averages to zero over angles . moreover , the configuration average of @xmath236 gives something which , when restricted to the hh and lh subspaces , is proportional to the identity matrix , so contributes only the scalar part of the scattering term . @xmath237 we separate the action of @xmath238 on the scalar and spin - dependent parts @xmath239 , and @xmath240 , of the wigner distribution @xmath241 . first on @xmath239 , which is written as @xmath242 , where @xmath243 is the anisotropic part @xmath244 averaged over impurities @xmath236 gives @xmath245 = \gamma^2_{h , l}\ , \openone , \end{array}\ ] ] the latter identity being valid because the matrix elements of the @xmath246 restricted to the hh and lh subspaces are proportional to the identity matrix . we also need to average @xmath247 , for which we note that @xmath248 for @xmath249 . averaged over impurities @xmath250 \displaystyle \gamma_s\ , \sigma_y \ , \gamma_s = & \displaystyle \frac{1}{4}\ , ( \gamma_y^2 - \gamma_x^2 - \gamma_z^2 ) \ , \sigma_y \\ [ 3ex ] \displaystyle \gamma_s\ , \sigma_z \ , \gamma_s = & \displaystyle \frac{1}{4}\ , ( \gamma_z^2 - \gamma_x^2 - \gamma_y^2 ) \ , \sigma_z . \end{array}\ ] ] this tells us that in the term @xmath251 only the average of @xmath240 , which we shall call @xmath252 , survives . then , writing @xmath253 @xmath254 looking at eq . ( [ eq : jn ] ) and ( [ eq : js ] ) we see that if we ignore the term linear in @xmath230 in each of them then they do not mix the scalar and spin distributions . we shall work for now in this approximation , which is justified because the terms omitted are higher order in the disorder potential . then we can write @xmath255 \displaystyle \frac{1}{\tau_\gamma } = & \displaystyle \frac{1 + \overline{\gamma_s^2}}{\tau}\\ [ 3ex ] \displaystyle \hat j_0 ( s ) + \hat j_m ( s ) = & \displaystyle \frac{\xi}{\tau } + \frac{\overline{\gamma_s^2}}{\tau } \ , ( \overline{s } + \xi ) - \frac{1}{\tau}\ , \overline{\gamma_s \ , \overline{s } \ , \gamma_s } . \end{array}\ ] ] the contribution to the scattering term due to the exchange splitting of the bands is	we formulate a theory of current - induced spin torques in inhomogeneous iii - v ferromagnetic semiconductors . the carrier spin-3/2 and large spin - orbit interaction , leading to spin non - conservation , introduce significant conceptual differences from spin torques in ferromagnetic metals . we determine the spin density in an electric field in the weak momentum scattering regime , demonstrating that the torque on the magnetization is intimately related to spin precession under the action of both the spin - orbit interaction and the exchange field characteristic of ferromagnetism . the spin polarization excited by the electric field is smaller than in ferromagnetic metals and , due to lack of angular momentum conservation , can not be expressed in a simple closed vectorial form . remarkably , scalar and spin - dependent scattering do not affect the result . we use our results to estimate the velocity of current - driven domain walls .
	for the two dimensional kinetic ising model at finite temperature , the local mean magnetisation @xmath0 , simply related to the fraction of time spent by a given spin in the positive direction , has a limiting distribution , singular at @xmath1 , the onsager spontaneous magnetization . the exponent of this singularity defines the persistence exponent @xmath2 . we also study first passage exponents associated to persistent large deviations of @xmath3 , and their temperature dependence . epsf _ submitted for publication to _ in this work we present a new approach to the study of persistence for systems undergoing phase ordering @xcite at finite temperature , which we shall illustrate on the case of the two dimensional ising model . in this approach persistence appears as a stationary property of the coarsening system , and the role of the onsager spontaneous magnetization at equilibrium @xmath4 is made apparent , thus revealing new fundamental features of phase ordering . it departs from previous approaches to finite temperature persistence @xcite , where these features did not appear . consider a system of ising spins @xmath5 located at sites @xmath6 , started from a random initial condition , and evolving under the heat bath dynamics at fixed temperature below the critical temperature . at each time step a spin is picked at random , and updated with the probability @xmath7 where the sum runs over the neighbours of site @xmath8 . under this dynamics spins thermalize in their local environment . therefore the system coarsens , i.e. domains of opposite signs grow and , in the scaling regime , the system is statistically self similar , with only one single characteristic length scale , which is the size of a typical domain @xcite . the question of persistence is to determine the fraction of space @xmath9 which remained in the same phase up to time @xmath10 @xcite ( or from time @xmath11 to time @xmath12 ) . for the two dimensional ising model at zero temperature , two phases coexist , corresponding to all spins equal to @xmath13 or all spins equal to @xmath14 . hence @xmath9 is equivalently defined as the fraction of spins which did not flip up to time @xmath10 @xcite , i.e. which were not swept by an interface between domains of all spins @xmath13 or all spins @xmath14 . numerical measurements indicate an algebraic decay @xmath15 , with @xmath16 @xcite . the definition of persistence at finite temperature below @xmath17 is more subtle to implement because one has to make clear what is meant by ` phase ' . by essence , in the coarsening process there is phase separation , each phase wanting to develop at the expense of the other . it is therefore intuitive that the system , though perpetually out of equilibrium , tries to reach locally one of the two equilibrium phases , corresponding to @xmath1 , where @xmath18 is the onsager spontaneous magnetization at equilibrium @xcite . hence in the scaling regime the average magnetization inside a domain measured on a scale of time small compared to the flipping time of the domain , should be close to the equilibrium magnetization at this temperature . coming back to persistence , the definition of @xmath9 should reflect , in one way or the other , the fact that a given point in space remained in a phase of average magnetization equal to @xmath1 up to time @xmath10 . this intuitive analysis is confirmed by what follows . the central point of our approach is to consider the statistics of the local mean magnetization simply related to the fraction of time spent by a spin in the positive direction in the limit of large times . this line of thought was already used in @xcite in the study of domain coarsening for the one dimensional ising model at zero temperature , or for the simple diffusion equation evolved from a random initial condition ( see also @xcite ) . the idea is that since persistence probes the past history of the system , a natural quantity to consider is @xmath19 , where @xmath20 is the spin at site @xmath8 and @xmath21 ( @xmath22 ) is the length of time spent by the spin pointing upward ( downward ) , with @xmath23 . the local mean magnetization is defined as @xmath24 for instance at zero temperature , the persistence probability @xmath9 is equal to @xmath25 since the event @xmath26 is identical to the event @xmath27 . since @xmath3 is a local quantity varying from site to site , one is naturally led to investigate the distribution of @xmath3 , @xmath28 for the one dimensional ising model at @xmath29 is was shown in @xcite by analytical arguments and numerical measurements that , when @xmath30 , @xmath31 converges to a limit distribution @xmath32 with density @xmath33 singular at @xmath34 with singularity exponent equal to @xmath35 . it is for instance easy to show that , when @xmath30 , the limit of @xmath36 is a constant equal to @xmath37 , the laplace transform of the two - time correlation with respect to the variable @xmath38 , at argument equal to 1 @xcite . this result therefore provides a _ stationary _ definition of persistence at zero temperature . the same holds for the diffusion equation @xcite . we now address the same questions at finite temperature . we first report on numerical results . = figure 1 depicts the histogram of the density of @xmath3 at time @xmath39 for values of @xmath40 ranging from @xmath41 to @xmath42 . already for such a short time , and for every temperature @xmath43 , the density is maximum around @xmath1 , the equilibrium magnetization ( [ eqons ] ) . at larger times and for @xmath44 , the density of @xmath3 becomes peaked around zero , i.e. the mean magnetization converges toward the average magnetization per spin @xmath45 , reflecting the fact that the system reaches equilibrium . at @xmath17 the peaking of the density of @xmath3 is observed to be very slow . finally at @xmath43 , @xmath31 converges , when @xmath30 , to a limit distribution @xmath32 with density @xmath46 given as in ( [ eqfm ] ) . the existence of a limit law at finite temperature @xmath47 relies on the same arguments as for the zero temperature case . for instance the convergence of @xmath48 to a constant equal to @xmath37 still holds since it only relies on the existence of a scaling regime @xcite . = the striking fact is that now the density concentrates on @xmath49 $ ] with an exponential decay with time of @xmath50 to @xmath41 . moreover the limit density is singular at @xmath1 , with a singularity exponent @xmath35 which defines the persistence exponent at temperature @xmath40 . this leads to the question of the temperature dependence of persistence for @xmath43 . the simplest assumption is that , during the coarsening process , the scales of time between a short time regime and the scaling regime decouple , yielding the following relation between the moments of the limit distributions at @xmath40 and at zero temperature , @xmath51 implying the identity of the limit distributions , if @xmath3 is rescaled by @xmath4 , and as a consequence , the temperature independence of @xmath2 . this would be in agreement with the usual view that zero temperature is an attracting fixed point for the dynamics of phase ordering @xcite . equation ( [ eqm2 t ] ) is hard to check by numerical measurements because the convergence of the data is observed to be slow . the difficulty is illustrated by figure 2 which depicts a plot of @xmath52 against the rescaled variable @xmath53 , at @xmath29 and @xmath54 , for @xmath55 , and at @xmath56 , for @xmath57 and @xmath58 . though one can not be conclusive on the sole basis of numerical measurements , data collapse seems nevertheless plausible . let us note that the limit distribution @xmath59 at @xmath29 is well approximated by a beta distribution , as was observed for the @xmath60 ising model @xcite , or for the diffusion equation @xcite . the singularity exponent of the beta distribution is found to be around @xmath61 . a more precise numerical determination of the exponent @xmath2 from the limit distribution of @xmath3 needs further work and will be presented elsewhere . let us summarize at this point . for @xmath43 , the local mean magnetization @xmath3 has a limit probability density when @xmath30 , defined on the interval @xmath49 $ ] , where @xmath4 is the equilibrium magnetization , and singular at both ends . this provides a _ stationary _ definition of persistence , which is a natural extension of the zero temperature case , where the singularity exponent defines the persistence exponent . these are the central results of the present work . = another new aspect of persistence introduced in @xcite is concerned with _ persistent large deviations_. the probability of persistent large deviations above the level @xmath62 ( @xmath63 ) , denoted by @xmath64 , is defined as the probability that the mean magnetization was , for all previous times , greater than @xmath62 @xcite , @xmath65 in other words one is interested in the persistence probability of the stochastic process @xmath66 @xcite . if one views the stochastic process @xmath20 as the successive steps of a fictitious random walker , then @xmath3 is the mean speed of the walker between 0 and @xmath10 , and @xmath64 is the probability that this mean speed remained larger than @xmath62 between 0 and @xmath10 . this probability is a natural generalization of the persistence probability @xmath67 , which corresponds for the walker to always stepping to the right . for the one dimensional ising model at zero temperature , @xmath64 was observed to decay algebraically at large times with an exponent @xmath68 continuously varying with @xmath62 @xcite . for @xmath69 , @xmath70 , the usual persistence exponent . figure 3 depicts a log - log plot of @xmath64 for the two dimensional ising model at zero temperature , @xmath62 varying from @xmath14 to @xmath71 , while figure 4 depicts the corresponding exponents @xmath68 , extracted from figure [ fig3 ] . let us mention that algebraic decay of @xmath64 was also observed for the diffusion equation @xcite , and that this quantity and the corresponding exponents @xmath68 can be exactly computed for the simple model considered in @xcite . = we now address the role of temperature for persistent large deviations . as is obvious from the first part of this work , if @xmath72 , then @xmath64 decays to zero exponentially rapidly . on the other hand , for @xmath73 one observes algebraic decay of @xmath64 , as at zero temperature . otherwise stated , @xmath74 separates two regimes of persistent large deviations , between exponential and algebraic . as a consequence , and by analogy with the zero temperature case , one could think of extracting the persistence exponent at finite temperature from the decay at large times of @xmath64 when @xmath75 , which leads to the formal definition @xmath76 however this definition is not easy to implement in practice . = we did not investigate the temperature dependence of @xmath68 in all generality . we restricted our study to the case @xmath77 which corresponds to @xmath78 we find that @xmath79 with @xmath80 for @xmath81 , after a crossover , @xmath82 takes the high temperature value @xmath83 while for @xmath43 it takes the low temperature value @xmath84 . ( see figure 5 . ) the explanation of the value of @xmath82 for @xmath85 is simple . since spins are independent , identifying as above @xmath20 , the spin at site @xmath8 , to the steps of a fictitious one dimensional symmetric random walker , @xmath86 represents the probability that the walker did not cross the origin up to time @xmath10 , which is , as is well known , decaying as @xmath87 @xcite . for decreasing temperatures , spins become more correlated , hence the exponent @xmath82 decreases . note that the first passage exponent @xmath82 is defined for @xmath88 , i.e. even in absence of coarsening . this work raises a number of questions . for instance , what is the temperature dependence of the two time correlation in the scaling regime , for @xmath43 ? is the hypothesis ( [ eqm2 t ] ) valid ? what is the behaviour of the distribution of @xmath3 at @xmath17 when @xmath30 ? at @xmath17 , is @xmath82 a new independent critical exponent , or is it related ( equal ? ) to the persistence exponent @xmath89 for the global magnetization @xcite ? let us mention that for the @xmath90 ising model the quantities studied here have similar behaviour . finally , in our view , an important point of the analysis presented here is that it may be applied to any coarsening system , since it relies mainly on scaling . we wish to thank j - p bouchaud , i dornic and j - m luck for interesting discussions .
accurate thermodynamic measurements are essential to understand fundamental properties of materials in various fields of physics . in condensed matter , the measurement of specific heat is a central characterization applicable to all kind of materials . low - temperature calorimetry is particularly suited for the investigation of superconductors and other novel systems with electronic phase transitions.@xcite such measurements require high resolution , since the electronic contribution to the heat capacity is only a minor part of the total heat capacity , except at the very lowest temperatures . the high resolution can be achieved through differential calorimeter designs and various temperature - modulated techniques.@xcite good accuracy is also often needed . the temperature dependence of the specific heat may , for instance , reveal central aspects of the nature of the electronic system , including energy gap structure , anisotropy , and possible signatures of quantum phase transitions . measurements are , furthermore , often performed in magnetic fields of various direction , using single crystals of highest available quality . this requires both a small calorimeter and small samples . all this together puts high demands on the calorimeter setup , typically excluding commercially available calorimeters . to meet this demand , calorimetry development is going in the direction of nanocalorimetry . nanocalorimetry is a rapidly growing area of research , driven by several fields of physics . nanocalorimeters include absorption sensors,@xcite devices to study transition enthalpies,@xcite fast scanning calorimeters to study microscopic nanostructure ensembles and thin films@xcite and kinetics and glass transitions of polymers@xcite , and combinatorial calorimeters.@xcite there are also several low - temperature microcalorimeters for @xmath12 samples@xcite and nanocalorimeters for heat capacity measurements of thin films@xcite and very small ( @xmath1 ) samples at high@xcite , intermediate,@xcite and down to low temperatures.@xcite going down in sample size favors the use of temperature - modulated techniques with corresponding high resolution , but makes it harder to obtain good absolute accuracy . this is partly due to an increasing relative contribution of the device addenda , but also due to non - adiabatic conditions and practical design issues , such as thermometry , system complexity , and thermal links considerations . devices for general use that combine high - resolution measurements of small samples with good absolute accuracy over an extended temperature range are hard to find . the membrane - based nanocalorimeter presented here is intended to fill this gap . our nanocalorimeter is developed for measurements of the temperature- and field dependence of the absolute specific heat of samples with a typical mass around @xmath13@xmath14 g , and for angular - dependent studies in magnetic fields . the device is built onto a pair of silicon nitride membranes using thin film techniques that provide low background heat capacity , less than @xmath3 at @xmath4 , decreasing to @xmath5 at @xmath6 , and a low thermal conductance , going from about @xmath15 at @xmath4 to @xmath16 at @xmath6 . the thermal relaxation time of the device is long enough ( ms to s range ) to enable ac steady - state and relaxation methods to be used concurrently . the high resolution of the ac steady - state method allows small changes in the heat capacity , such as contributions from the electronic specific heat , to be accurately determined , and investigations of phase transitions and phase diagrams to be performed . absolute accuracy is obtained by a combination of low background addenda , a stacked calorimeter design , and extensive measurement electronics operated with self - regulation and frequency feedback . the compact format enables the calorimeter to be placed on sample holders for rotation in magnetic field . the calorimeter could even be used for studies of dynamic ( frequency dependent ) heat capacity.@xcite the device is thus a versatile tool for general thermodynamic studies of small samples . ( a ) cross - sectional schematic of the calorimeter on top of a copper base with vacuum channels . ( b ) top view layout of the calorimeter with the two membrane - based calorimeter cells surrounded by 20 bonding pads that connect the calorimeter to the measurement electronics . ( c ) schematic of one of the @xmath17 membrane cells , composed by ac heater , thin film geau thermometer , and offset heater . the sample is placed on the @xmath18 central thermometer area . on top of the stack , a thermalization layer made of au is deposited if smaller samples are to be used , to obtain good internal thermalization and a uniform temperature distribution over the whole thermometer . ( d ) illustration of the active layers . all active layers are electrically insulated from each other and the sample by sio@xmath19/alo@xmath20 layers ( not shown ) . ] the nanocalorimeter is built on top of two custom - designed , pre - fabricated silicon nitride membranes , @xmath21 in size and @xmath22 thick ( spi supplies , west chester , usa ) . the membranes are suspended by a si frame @xmath23 , which is attached by means of stycast to a copper base as illustrated in fig . [ fig1](a ) . figure [ fig1](b ) shows a top view of the calorimeter layout . each cell , shown in fig . [ fig1](c ) , is composed of a stack of ac heater , thermometer , and offset heater in the central area of the membrane . between each of these active layers there are electrical insulation layers . all layers are fabricated using photolithography , deposition ( e - beam evaporation or sputtering ) , and double - layer resist lift - off . the active layers are illustrated separately in fig . [ fig1](d ) . the ac heater , shown in fig . [ fig1](d ) , is a meander - shaped resistor made of e - beam evaporated titanium , @xmath24 wide and @xmath25 thick , which covers the central sample area . it is used to oscillate the sample temperature with a well - defined ac power . it has a four - point probe geometry to allow accurate determination of the power at the sample without contributions from lead and contact resistances . ti becomes superconducting below about @xmath26 , but since the superconductivity can be suppressed by relatively modest fields the heaters may still be used at even lower temperatures . if needed , the film thickness could also be decreased to suppress @xmath27 . the active part of the thermometer , shown in fig . [ fig1](d ) , is a @xmath28 square made of @xmath29 thick , sputtered @xmath7 alloy . it senses the sample area using a four - point probe configuration and is in direct thermal contact with the sample , thus probing the actual sample temperature . consequently , heat dissipation in the thermometer does not pose a problem after initial calibration so that the thermometer can be operated at rather high powers to achieve a high sensitivity . furthermore , there is no need to rely on measurements of the frame / base temperature , unlike when thermocouples are used.@xcite this eliminates hysteretic effects when sweeping temperature and results in a high reproducibility . the sensor layer is fabricated using rf magnetron sputtering from a 2-inch target made of cast @xmath7 with nominally 17 at% au.@xcite the chip is annealed at @xmath30 on a hotplate for at least 1 hour after deposition , resulting in a room temperature resistivity @xmath31 and dimensionless sensitivity @xmath32 between @xmath4 and @xmath33 , increasing to about @xmath34 at lower temperatures . the geau sensor layer is deposited on top of au leads as shown in fig . [ fig1](d ) . the leads are in turn connected to external leads in ti , as seen in fig . [ fig1](c ) . by combining metals with high ( au ) and low ( ti ) thermal conductivity a more well - defined isothermal area in the center of the membrane is obtained.@xcite the offset heater , shown in fig . [ fig1](d ) , is driven by a dc current and can locally increase the sample temperature up to at least @xmath35 above the base temperature . this heater is designed to give an isothermal interior area , but is not in direct contact with the active part of the thermometer . it is thus less suitable as an ac heater , since the thermal diffusion between thermometer and heater would have to be taken into account . titanium has good robustness so the deposited layer can be quite thin . its addition to the background heat capacity is rather insignificant , but to simplify the fabrication process the layer is sometimes skipped . the leads that connect the active layers to the external bonding pads are also in ti . because of the relatively low thermal conductance of ti compared to many other metals , the use of ti for the leads ensures a long relaxation time dominated by the membrane itself . outside the membrane area , a thick layer of au is deposited onto the leads to minimize lead resistance effects and to provide a suitable layer for bonding , see fig . [ fig1](a ) . as a last element of the calorimeter stack , a @xmath36 square may be deposited as a thermalization layer to distribute the temperature evenly over the sample area if small samples are to be studied . this layer may be deposited simultaneously with the outer au bonding pad layer . the electrical insulation layers , not shown in fig . [ fig1 ] , are made by thermally evaporated aluminum oxide in combination with sputtered sio@xmath19 . the sio@xmath19 is deposited directly following the deposition of heaters or geau layer , while the alo@xmath20 layers are patterned as separate layers . there are in total three alo@xmath20 layers in the shape of rounded squares of different sizes : those between heaters and thermometer measure @xmath37 while the last insulation between thermometer and thermalization layer is @xmath38 . care was taken to design the layers so that edges of different layers do not coincide . this greatly reduces the risk of shorts between layers , especially for thin insulation layers . ( a ) calorimeter bonded onto cryostat plug - in . ( b ) calorimeter cell with a sample covering ac heater and thermometer . the membrane and electrical insulation layers appear transparent . this particular calorimeter has no offset heater . ( c ) microscope image of the central part of a calorimeter cell , illustrating the active layers and insulation ( but without geau layer for clarity ) . ] figure [ fig2](a ) shows a picture of the calorimeter bonded onto a cryostat sample holder plug - in . the calorimeter requires up to 20 wires , but can otherwise be fitted onto most cryogenic sample holders . the sample is placed on one of the membrane cells by means of a simple micro - manipulator , or , with some practice , by hand . the other cell may either carry a reference sample or be left empty . figure [ fig2](b ) shows a calorimeter cell with a typical sample ( the pb sample discussed in section [ sec : meas ] ) . the central parts of the calorimeter are shown in fig . [ fig2](c ) , where also the larger electrical insulation layers can be seen . in the standard measurement mode , known ac and dc currents flow through the heaters and thermometers . to practically enable the measurements of temperatures and oscillation amplitudes , a set of time and phase synchronized lock - in amplifiers based on a field - programmable gate array ( fpga ) is used.@xcite we implemented such an instrument using the pxi-7854r card by national instruments . the card has integrated @xmath39 adcs and @xmath40 dacs , providing eight integrated , simultaneous - sampling analog inputs and outputs for ac and dc biasing . for each input , the first and/or second harmonic amplitudes with corresponding phases are extracted , as well as the dc component . a central phase generator delivers the digital reference for all inputs and outputs . the resulting instrument thus allows tuning of output voltages and working frequency during the measurements , without any loss of correlation between inputs and outputs . each signal , before being read , passes through a low - noise , custom - built preamplifier stage . in total eleven preamplifiers are used , eight of which have variable gain ( between 1 and 5000 ) , controlled by the fpga lock - in , the others with fixed gain ( 100 or 1000 ) . while measuring , the eight variable - gain preamplifiers are automatically adjusted to maximize their performance . the circuit scheme used for the current bias and voltage read - out of the thermometers is illustrated in fig . an automated process is implemented that auto - adjusts the current through the sample - side thermometer in order to regulate the voltage across the thermometer . the applied voltage @xmath41 is thus varied to keep the dc component @xmath42 of the sample voltage @xmath43 equal to a variable setpoint value that is typically around @xmath44 and almost constant over the full temperature range . when the resistance of the thermometer increases with decreasing temperature , the applied current decreases and a suitable power is still delivered . the reading of @xmath42 is made synchronously by the lock - in so that frequency - dependent signals are separated from the dc component . changes to @xmath41 are synchronized with the measurements as well . thermometer bias and read - out circuit . the sample and reference geau thermometer resistances @xmath45 and @xmath46 are measured in four - probe configurations with the current provided by synchronous voltage sources @xmath41 and @xmath47 . the currents through the thermometers are measured by the voltages over @xmath48 series resistors . optional , adjustable voltage dividers are used to avoid any electrical discharge from destroying the membranes during initial connection . ] since the sample and reference thermometers are fabricated together at the same time , they are quite well balanced and display the same temperature dependences , within measurement uncertainty . there may , however , still be a small imbalance in the resistance ratio of the order of 1% between the two sides , arising from the lithographic tolerance . to compensate for this , a reference adjustment system is applied to vary the reference output @xmath47 in fig . [ fig3 ] to keep @xmath49 so that @xmath50 , rather than setting the currents or powers equal on the two sides . by balancing the thermometers in this way , the time - varying temperature difference between sample and reference is given by the @xmath51 signal through the simple relation @xmath52 here @xmath53 is the absolute temperature , which is obtained from the dc component of the thermometer resistance . the additional temperature difference caused by the power difference is usually insignificant , but can be compensated for by a corresponding power from the offset heater . the absence of dc offsets in the bridge differential @xmath51 makes it possible to apply a high amplification to @xmath51 without overloading , which results in highest possible resolution . the close proximity of sample and reference thermometers further eliminates common noise sources such as electromagnetic interference and temperature variations of the base frame . it should be noted that eq . ( [ eqdeltatemp ] ) has to be modified when the thermometer resistance becomes comparable to the series resistance of fig . [ fig3 ] , to include the effect of a varying bias current . in the ac steady - state method @xcite the sample temperature is made to oscillate with an amplitude typically in the range @xmath54@xmath55 . this modulation is created by an ac power , which for resistive heating is given by @xmath56 . it is thus generated by an ac current with rms amplitude @xmath57 and angular frequency @xmath58 flowing through the ac heater resistor @xmath59 . the temperature response of the cell is given by @xmath60 . @xmath61 is the base temperature , @xmath62 is the dc offset due to the time - averaged power supplied by the heater resistance , where @xmath63 is the thermal conductance between sample and thermal bath ( si frame ) , and @xmath64 is the oscillating term whose steady - state amplitude @xmath65 is directly related to the heat capacity of the sample : @xmath66^{-1/2}.\ ] ] here @xmath67 is the external relaxation time between sample and external thermal bath and @xmath68 is a rather complicated function of the internal relaxation time between sample and calorimetric cell , @xmath69 , where @xmath70 is the sample heat capacity and @xmath71 is the thermal conductance between the sample and the central cell platform . the value of @xmath71 strongly depends on the agent used to attach the sample to the nanocalorimetric cell . equation ( [ eqtac ] ) is impractical to use to obtain @xmath72 from @xmath65 , due to the complications of @xmath68 . we have , however , previously shown@xcite that for a system with good thermal connection between the active layers , but with possibly significant @xmath73 , the thermometer temperature oscillation @xmath74 and corresponding phase @xmath75 between power and temperature oscillation can be found as [ eqtphi ] t _ & = + & = provided that @xmath72 and @xmath76 are taken as c & = c _ + ( 1-g ) c _ + k & = k_+g k _ + g & = [ eq_cktotal ] here , @xmath77 is the empty cell contribution to the heat capacity and @xmath78 is the effective thermal link of the membrane.@xcite equation ( [ eqtphi ] ) can be wrapped around into the practical , functional relations [ eqck ] c & = + k & = which form the basis of evaluating the measurements . note that the phase is carrying information that is needed to achieve good absolute accuracy . the phase can be used to verify that the frequency is selected correctly , i.e. , that @xmath79 so that @xmath80 . if the frequency is so high that @xmath81 in eq . ( [ eqck ] ) , it is very likely that @xmath82 contributes significantly in eq . ( [ eq_cktotal ] ) . a known phase is also necessary to accurately extract @xmath63 from ac steady - state measurements . when using the ac steady - state technique in practical terms , we apply an ac current to the heater(s ) and measure the thermometer and series resistance voltages @xmath43 , @xmath83 , @xmath84 , @xmath85 , and @xmath51 , defined in fig . [ fig3 ] , at the second harmonic of the heater current frequency ( simultaneously with the synchronous dc mean ) . the amplitudes of the sample and reference temperature oscillations @xmath86 and @xmath87 , are related to the corresponding measured voltages over the thermometers and currents through the reference resistors by @xmath88 here both @xmath53 and sensitivity @xmath89 are obtained from the dc measurement of the thermometer resistance @xmath90 and previous calibration of @xmath91 . for the temperature differential the corresponding expression is @xmath92 where @xmath93 and @xmath94 is the series resistance for current measurement . all ac signals in eqs . ( [ tacfromuac ] ) and ( [ tdifffromudiff ] ) refer to amplitudes in the steady state . note that @xmath95 in general , since the sample and reference temperature oscillations may have different phases . in the simplest measurement case , the reference heater power is kept off . this method is used when the heat capacity of the sample is fairly large as compared to the heat capacity of the empty cell and no reference sample is placed on the reference side . the sample temperature oscillation is then measured through the differential signal @xmath96 while @xmath97 . since @xmath51 has no dc offset , the differential signal yields a significantly higher resolution than @xmath98 . to subtract @xmath77 from @xmath72 , sample and empty cell are measured in separate runs . the measurement of @xmath65 is thus made differentially , but the measurement of @xmath72 is not . if the sample is small , or if a similar - size reference sample is added to the reference side , a truly differential measurement mode can be employed . in this case , the same power is applied to both sample and reference sides ( @xmath99 ) . the heat capacities @xmath100 and @xmath101 of the individual sides are still given by eq . ( [ eqck ] ) , but the differential heat capacity @xmath102 can be obtained as well from @xmath103 : @xmath104 equation ( [ c_diff ] ) is valid under the assumption that @xmath76 is the same for both sample and reference sides . since @xmath63 is given mainly by the membrane itself , and other contributing layers are manufactured lithographically , this requirement is easy to fulfill in normal cases . if a large sample is studied , the sample itself may enhance @xmath76 through a small , but nonzero @xmath82 . the absolute accuracy of eq . ( [ c_diff ] ) may in this case be tested by a separate verification measurement with @xmath105 . if the differential heat capacity mode is used with a large sample but without reference , the signal from the reference side will dominate @xmath51 , and the benefit of a differential measurement would be lost . an alternative to reverting to a single - side measurement can in this case be to adjust the amplitude and output phase of the reference heater power to maintain @xmath106 . this mode of measurement is , however , still unexplored . we have previously shown@xcite that good absolute accuracy can be obtained provided that one has good control of the phase in eq . ( [ eqck ] ) . in fig . [ fig4 ] , the frequency dependence of @xmath86 and @xmath75 is shown for a typical sample . measured frequency dependence of temperature oscillation amplitude @xmath65 and phase @xmath75 , expressed as @xmath107 and @xmath108 , respectively , with @xmath109 and @xmath72 constant ( @xmath110 ) . at low frequency @xmath111 and measurements yield good absolute accuracy . at too low frequencies , however , the signal no longer comes from the heat capacity but from the thermal link , and the resolution decreases . note that the middle of the adiabatic plateau ( where @xmath112 is constant ) is not corresponding to the best measurement frequency if good accuracy is required . ] good accuracy is found at low frequencies , where @xmath111 . at higher frequencies , the effect of @xmath113 is no longer negligible and the accuracy is quickly deteriorating . for the resolution , the conditions are the opposite ; at high frequencies the resolution is good , but at low frequencies , @xmath114 , eq . ( [ eqtphi ] ) is reduced to @xmath115 and heat capacity is no longer probed . by fixing the phase @xmath75 to a constant value in the range where measurements yield both good absolute accuracy and good resolution , optimal conditions are found . we do this by continuously adjusting the frequency @xmath116 during the measurement to keep @xmath108 constant by means of an auto - tuning routine in the fpga lock - in . by having a control loop time equal to a single sample of the adc , the frequency is smoothly adjusted without introducing additional noise to the measurements . the frequency is thus varying even during a single cycle of the output . this is possible thanks to the synchronous sampling . it should be noted that the middle of the adiabatic plateau ( with @xmath117 ) is typically at too high frequencies for good absolute accuracy , as seen in fig . [ fig4 ] . the thermal relaxation method @xcite consists of applying a known power to the heater to raise the sample temperature an amount @xmath118 above the base temperature @xmath61 . after a stable sample temperature has been reached , the heater power is turned off and the temperature is allowed to relax back to @xmath61 . the time dependence of the relaxation is exponential and depends on the external time constant of the system : @xmath119 when using the thermal relaxation method , we apply a square wave with sufficiently low repetition rate to the sample heater , while keeping the reference heater off . the induced temperature response is directly given by eq . ( [ eqdeltatemp ] ) , from which @xmath118 and @xmath120 can be obtained . the thermal conductance @xmath63 between sample and base frame is then found as @xmath121 where @xmath122 is given by the directly measured step in heater power . the sample heat capacity including cell addenda is finally given by @xmath120 and @xmath63 , according to @xmath123 to practically perform relaxation measurements , a routine was developed that automatically acquires the voltage pulses , fits the exponential decay to data within 10% and 90% of the relaxation , and determines @xmath118 . in this way , the heat capacity and device thermal conductance can be measured as a function of temperature or magnetic field with immediate results displayed on the screen . ( a ) temperature dependence of the geau thermometer resistance and dimensionless sensitivity @xmath124 . ( b ) thermometer magnetoresistance expressed as apparent relative temperature change in a magnetic field of @xmath125 . the inset shows the corresponding field dependence at different temperatures . the magnetoresistance is following a parabolic field dependence except for temperatures below @xmath126 at high fields.,title="fig : " ] + ( a ) temperature dependence of the geau thermometer resistance and dimensionless sensitivity @xmath124 . ( b ) thermometer magnetoresistance expressed as apparent relative temperature change in a magnetic field of @xmath125 . the inset shows the corresponding field dependence at different temperatures . the magnetoresistance is following a parabolic field dependence except for temperatures below @xmath126 at high fields.,title="fig : " ] before any measurements , the geau thermometers on sample and reference sides need to be calibrated against a known thermometer . we use a cernox sensor on the sample holder . the calibration is done in the presence of a few mbar helium gas to increase the thermal link between geau thermometers and si base frame , which quickly reduces @xmath120 and temperature offsets due to self - heating . the typical temperature dependence of the geau thermometer resistance is shown in fig . [ fig5](a ) . once the calibration curve has been obtained , one can attain the temperature dependence of the sensitivity @xmath89 which is approximately constant from @xmath4 down to about @xmath127 , below which it increases slowly as shown in fig . [ fig5](a ) . the temperature dependence of the sample to reference thermometer resistance ratio @xmath128 is constant within experimental uncertainty over the entire temperature range . the same calibration curve can thus be used for both sides , except for a scaling pre - factor . thermometers deposited in the same sputtering cycle display very similar temperature dependences , while thermometers fabricated at different times require individual calibrations , although differences are small enough for a standard calibration curve to be used initially in most cases . the thermometer resistance is stable over time provided that the thermometer is not heated excessively after the heat treatment at @xmath129 . to obtain @xmath89 and @xmath53 from @xmath90 while avoiding numerical noise and other artifacts , we fit an analytical expression to the calibration data . the thermometer @xmath130 relation is well described by @xmath131 here @xmath132 , @xmath133 , and @xmath134 are constants . remaining small deviations are fitted by a high - degree polynomial in @xmath135 . an analytical expression of @xmath89 is then directly found from the fit parameters , and is given by @xmath136,\ ] ] up to the contribution from the polynomial correction . this procedure to obtain temperature and sensitivity from resistance avoids interpolations , which , even in log - log scale , tend to give rise to artificial kinks that may become significant in certain cases such as when studying specific heat differences . resistive thermometers require corrections for magnetic - field - induced changes at low temperature . the magnetoresistance of the geau sensor was studied by sweeping a magnetic field between @xmath137 and @xmath138 at several fixed temperatures from @xmath139 to @xmath140 . the relative change in apparent temperature was then calculated as @xmath141 , where @xmath89 is the zero field sensitivity . the curves , shown in the inset of fig . [ fig5](b ) , can be fairly well approximated by a parabolic field dependence @xmath142 above @xmath143 . the magnetoresistance is positive and its magnitude decreases quite quickly with increasing temperature . the field - induced error at @xmath138 , if no correction is made , is @xmath144 at @xmath145 and @xmath146 at @xmath140 , as shown in fig . [ fig5](b ) . the nanocalorimeter was characterized thoroughly from room temperature down to @xmath2 . figure [ fig6](a ) shows the heat capacity of the empty cell as obtained from eq . ( [ eqck ] ) . ( a ) empty cell heat capacity @xmath147 as a function of temperature . ( b ) device thermal conductance @xmath63 . the insets show the low - temperature behavior in log - log scale . note that @xmath147 and @xmath63 display a fairly similar temperature dependence , resulting in a rather constant time constant @xmath148 as a function of temperature for the empty device.,title="fig : " ] + ( a ) empty cell heat capacity @xmath147 as a function of temperature . ( b ) device thermal conductance @xmath63 . the insets show the low - temperature behavior in log - log scale . note that @xmath147 and @xmath63 display a fairly similar temperature dependence , resulting in a rather constant time constant @xmath148 as a function of temperature for the empty device.,title="fig : " ] this heat capacity will , under certain conditions , depend on the choice of woking frequency ( i.e. @xmath108 ) , since the amount of membrane that is temperature - modulated decreases with increasing frequency.@xcite the characteristic frequency of the membrane is , however , typically higher than the frequency of normal measurements , making it possible to treat @xmath147 as a reproducible background addenda for a given calorimeter and to assume that @xmath77 is frequency independent for large samples ( i.e. with @xmath149 ) . in differential mode , the addenda heat capacity is normally less than 5% of @xmath147 . a typical noise level is @xmath150 at @xmath127 . the noise level expressed as @xmath151 is fairly constant over the entire temperature range . with a typical measurement time of @xmath152 and a moderately large @xmath153 , it is possible to reach @xmath154 . the thermal link @xmath63 of the calorimeter cell is shown in fig . [ fig6](b ) . it can be obtained from eq . ( [ eqck ] ) provided that the measurements are made at low enough frequency , @xmath155 , or through relaxation measurements . while the cell heat capacity is dominated by the membrane that has a fairly high debye temperature , the thermal link is given by a combination of membrane and metallic leads . the characteristic time constant @xmath156 is nevertheless fairly temperature independent . figure [ fig7](a ) shows the range of operational power of the ac heater as a function of temperature . the power required for a certain ratio @xmath153 between temperature oscillation amplitude and absolute temperature is not depending on @xmath72 but only @xmath63 , provided that the phase @xmath75 is kept constant . an easy way to adjust the power to the proper level is thus to maintain a constant ratio @xmath157 while adjusting the frequency so that @xmath75 is constant . in this way , the temperature offset due to the ac heater power is always a constant fraction of @xmath53 as well . to maintain a similar power for the dc bias of the thermometer , the @xmath42 should decrease from about @xmath44 at @xmath35 to @xmath158 at @xmath6 . in practice , @xmath42 can be kept almost constant , so that the relative temperature offset due to thermometer self - heating is somewhat higher at the lowest temperatures and somewhat lower at the highest . ( a ) range of operational ac heater power as a function of temperature for measurement frequency adjusted so that @xmath159 . ( b ) typical range of working frequency @xmath160 as a function of temperature , with and without sample . note that the sample specific heat controls the frequency range over which good absolute accuracy is obtained . different samples thus require different frequency adjustment.,title="fig : " ] + ( a ) range of operational ac heater power as a function of temperature for measurement frequency adjusted so that @xmath159 . ( b ) typical range of working frequency @xmath160 as a function of temperature , with and without sample . note that the sample specific heat controls the frequency range over which good absolute accuracy is obtained . different samples thus require different frequency adjustment.,title="fig : " ] figure [ fig7](b ) shows the typical range of working frequency @xmath160 as a function of temperature , with and without sample . for the empty device , the time constant @xmath148 is rather temperature independent , leading to a fairly constant @xmath160 . with a sample , the frequency @xmath161 is lowered , but the frequency also becomes more strongly temperature dependent . this variation depends on the sample heat capacity . it is thus clear that only narrow temperature ranges can be studied with combined good accuracy and resolution if a constant frequency is used as in the case of traditional ac calorimetry . as an initial test of the calorimeter , we measured a small gold sample from @xmath162 down to about @xmath33 to compare the relaxation and ac steady - state techniques . the temperature dependence of the heat capacity , shown in fig . [ fig8 ] , was also compared with available data to study the absolute accuracy . measured heat capacity versus temperature of a small gold grain . the contribution of the membrane heat capacity has been subtracted . there is also some contribution from the apiezon - n grease used to attach the sample . this addenda has not been accounted for , but is expected to be of the order of @xmath163@xmath164 . the heat capacity measured by geballe and giauque@xcite on a sample more than @xmath165 times larger is scaled to agree with our data at @xmath166 . ] it is seen that the over - all agreement between the two measurement methods is fairly good . the ac steady - state curve , however , lies somewhat above the relaxation curve for temperatures below @xmath127 . the heat capacity measurement by geballe and giauque@xcite on roughly @xmath167 ( ! ) au was scaled to agree with our data at @xmath166 , where the relaxation and ac steady - state measurements coincide . the relaxation curve follows the temperature dependence of the literature with deviations within 5% over the full temperature range . from the scaling , the sample mass is estimated to @xmath168 , which lies within the uncertainty of the volumetric measurement of the sample size , initially estimated to be @xmath169 using a simple microscope . the real sample mass is likely in between the two numbers , since the data of fig . [ fig8 ] includes a small contribution from the apiezon - n grease that was used to attach the sample . to illustrate the low - temperature capabilities of the calorimeter , we studied the heat capacity of a @xmath11 , 99.999@xmath170 pure pb sample , cut from a single crystal , as a function of temperature and in magnetic fields . figure [ fig9](a ) shows the measured specific heat in the superconducting and normal states , plotted as @xmath171 vs @xmath53 . ( a ) temperature dependence of the low - temperature specific heat expressed as @xmath171 of pb in the superconducting and normal state ( obtained by a @xmath172 field ) . the inset shows the measured heat capacity at higher temperatures before and after subtraction of addenda , and the contributions from membrane and apiezon - n grease . ( b ) specific heat difference @xmath173 , with the data provided by neighbor _ _ etal.__@xcite for comparison . the inset shows the entropy difference @xmath174 obtained by integrating @xmath175.,title="fig : " ] + ( a ) temperature dependence of the low - temperature specific heat expressed as @xmath171 of pb in the superconducting and normal state ( obtained by a @xmath172 field ) . the inset shows the measured heat capacity at higher temperatures before and after subtraction of addenda , and the contributions from membrane and apiezon - n grease . ( b ) specific heat difference @xmath173 , with the data provided by neighbor _ _ etal.__@xcite for comparison . the inset shows the entropy difference @xmath174 obtained by integrating @xmath175.,title="fig : " ] the specific heat jump in zero field at the transition temperature @xmath176 is seen clearly . the normal state curve was obtained by applying a @xmath172 magnetic field to suppress the superconductivity . the inset shows the heat capacity at higher temperatures . the addenda heat capacity from membrane and grease were measured in separate runs ( i.e. not in true differential mode ) . apiezon - n is typically used as a thermal contact agent for low - temperature experiments , but undergoes a glass transition at above @xmath177 , which leads to a somewhat irreproducible high - temperature addenda,@xcite decreasing the absolute accuracy at high temperatures . the membrane cell dominates the addenda at the absolute lowest temperatures , but already at about @xmath126 the apiezon contribution becomes the main background . the membrane and grease addenda are subtracted in the main figure [ fig9 ] . after subtraction , the superconducting state measurements still display a @xmath178 residual gamma term , i.e. , a remaining linear - in-@xmath53 contribution to the specific heat . this could possibly be due to an incompletely accounted background addenda . the ratio @xmath179 at @xmath27 is , however , only @xmath180 of expected . we therefore believe that the unaccounted addenda is a non - superconducting part of the sample entering @xmath72 but not @xmath181 , possibly arising from an oxidized surface layer , or from vacancies and dislocations that were not annealed away before the measurements . figure [ fig9](b ) shows the temperature dependence of the specific heat difference @xmath173 . this difference is insensitive to background addenda , which makes it a good probe of accuracy and reproducibility . from the temperature dependence , fundamental properties such as the superconducting gap energy and coupling strength can be obtained.@xcite the measurements provide a sommerfeld parameter @xmath182 and @xmath183 in good agreement with literature.@xcite the temperature dependence of @xmath175 is also following the expected behavior , as seen by comparing the measurements with polynomial - fit data provided by neighbor _ _ etal.__@xcite while the resolution remains good at all temperatures , the accuracy decreases somewhat at @xmath184 , which can be seen in fig . [ fig9](b ) as some wiggles in @xmath175 . this can be attributed to the difficulties in obtaining an accurate calibration and corresponding sensitivity of the thermometer using @xmath185 power levels . one way to overcome this problem may be to calibrate the thermometer simultaneously with the measurements , with a self - consistency requirement on the thermal link . such a method has been successfully tested , but requires further investigation . integrating @xmath175 gives the entropy difference @xmath186 , shown in the inset of fig . [ fig9](b ) . the entropy - conservation requirement is fulfilled within a @xmath187 uncertainty of @xmath188 , obtained from the low - temperature slope of @xmath189 . the free energy difference @xmath190 is then obtained as @xmath191 , where @xmath192 is found by integrating @xmath181 from @xmath27 to @xmath53 . from @xmath190 , the thermodynamic critical field @xmath193 is calculated from the relation @xmath194 , where @xmath195 is the molar volume ( or sample volume , if @xmath190 is given in units of energy ) . figure [ fig10](a ) shows @xmath193 obtained in this way using the data in fig . [ fig9](b ) . ( a ) thermodynamic critical field @xmath193 . the small , closely spaced symbols correspond to @xmath193 as obtained from the @xmath196 data of fig . [ fig9](b ) . they display the expected@xcite small deviation from the two - fluid expression , which is shown for reference . also shown is the directly measured location of the superconducting transition in various magnetic fields ( big circles ) . the inset shows normalized heat capacity measured on increasing fields at different temperatures . ( b ) temperature dependence of the specific heat in small magnetic fields near @xmath27 . ( c ) transition in a @xmath197 magnetic field measured with different @xmath65 . while improving the energy resolution , too high @xmath65 results in a pronounced @xmath53 smearing . ] the temperature dependence of @xmath198 for pb is expected to display a small , positive deviation from the two - fluid expression @xmath199 $ ] . a weak - coupling bcs superconductor , on the other hand , would display a negative deviation.@xcite the difference between @xmath193 , as obtained from the measured @xmath196 , and the two - fluid expression is clearly seen in fig . [ fig10](a ) . indeed , the temperature dependence of the deviations themselves are within a few % of the deviations obtained by decker _ _ etal.__@xcite the most uncertain factor in going from heat capacity to specific heat in nanocalorimetry is the determination of sample mass . it can be done through a careful measurement of volume and density , or from a known reference point at some temperature , such as room temperature or @xmath27 . since a microscopic volume measurement would require sub-@xmath200 resolution , which is difficult for soft materials such as pb , we used the measurement of @xmath72 around @xmath27 by shiffman _ _ etal.__@xcite to obtain the scale in fig . [ fig9 ] . for type - i superconductors , it is however also possible to directly measure @xmath193 by studying the superconducting transition in magnetic field . such measurements at various temperatures and magnetic fields are shown in the inset of fig . [ fig10](a ) and in fig . [ fig10](b ) . the resulting @xmath193 and @xmath201 transitions , shown as big , green circles in the main panel of fig . [ fig10](a ) , agree well with @xmath193 as obtained from the measurement of @xmath181 . the sample volume can thus be found directly from a comparison of @xmath190 ( measured in units of energy ) and @xmath202 ( having units of energy per volume ) . the measurements in magnetic field were made with @xmath203 perpendicular to the plate - like sample . ( the sample is shown in fig . [ fig2 ] ) . this causes @xmath72 to become field dependent in the superconducting state , due to the large demagnetization factor that drives the sample into the intermediate state . the effect is clearly seen in the field - dependence curves of the inset of fig . [ fig10](a ) . at low fields the sample is in the meissner state and @xmath70 is constant , but at higher fields @xmath70 starts to increase . one could interpret this effect as a distributed latent heat when normal domains enter the superconductor . in fig . [ fig10](b ) it is seen that the specific heat in small magnetic fields is higher than the zero field specific heat near @xmath198 . while the sample temperature oscillates , a small fraction of the sample is undergoing the transition back and forth between the meissner and normal states , with accompanying latent heat @xmath204 . this causes the @xmath65 amplitude to decrease , making the latent heat appear as an excess specific heat @xmath205 . the excess specific heat thus relates to the fraction of the sample that undergoes the superconducting transition at each temperature . it is tempting to quantify the latent heat from such measurements . however , only a fraction of the total latent heat is found in this way . this is due to possible hysteretic effects of the first - order transition in combination with an incomplete analysis of the temperature oscillation , which will display higher harmonics when latent heat is involved . as a final illustration of the capability of the calorimeter , the transition in a @xmath197 magnetic field is shown in fig . [ fig10](c ) for different temperature oscillation amplitudes @xmath65 . the sharp latent heat peak , which is not present in zero field [ cf . [ fig10](b ) ] , is a good probe of combined high resolution of both temperature and specific heat . the peak is only seen if @xmath65 is small enough . by increasing @xmath65 the specific heat resolution will increase , but the transition is then quickly smeared out . the latent heat involved in this transition is of the order of a few @xmath206 . in summary , we have developed a membrane - based nanocalorimeter for specific heat measurements of small samples and thin films over an extended temperature range from above room temperature down to below @xmath6 . our device has sub - pj / k resolution at low temperature , corresponding to @xmath207 for thin films and @xmath208 heat exchanges , and is capable of probing @xmath1-sized samples with combined high resolution and good absolute accuracy , thus exceeding the typical capability of commercial calorimeters by almost four orders in sample size . the calorimeter features a differential design with a variable - frequency technique where the measurement conditions are automatically maintained at optimal conditions . the versatility of the calorimeter invites the exploration of several novel ac measurement procedures in addition to the ones described here , including power - compensation and multi - frequency modes . the ultimate capability of the calorimeter is thus still an open question . we thank s. latos and p. a. favuzzi for assistance with calorimetry development and r. nilsson for contributing to the fpga - based lock - in amplifier . initial development of the nanocalorimeter was performed at argonne national laboratory in collaboration with u. welp , w .- k . kwok , and g. w. crabtree . financial support from the swedish research council and technical support from the su - core facility in nanotechnology is gratefully acknowledged . a. rydh , _ calorimetry of sub - microgram grains _ , in encyclopedia of materials : science and technology , 2006 online update , k. h. j. buschow , m. c. flemings , r. w. cahn , p. veyssire , e. j. kramer and s. mahajan ( eds ) . elsevier ltd . , oxford , 2006 .	a differential , membrane - based nanocalorimeter for general specific heat studies of very small samples , ranging from @xmath0 to sub-@xmath1 in mass , is described . the calorimeter operates over the temperature range from above room temperature down to @xmath2 . it consists of a pair of cells , each of which is a stack of heaters and thermometer in the center of a silicon nitride membrane , in total giving a background heat capacity less than @xmath3 at @xmath4 , decreasing to @xmath5 at @xmath6 . the device has several distinctive features : i ) the resistive thermometer , made of a @xmath7 alloy , displays a high dimensionless sensitivity @xmath8 over the entire temperature range . ii ) the sample is placed in direct contact with the thermometer , which is allowed to self - heat . the thermometer can thus be operated at high dc current to increase the resolution . iii ) data are acquired with a set of eight synchronized lock - in amplifiers measuring dc , @xmath9 and @xmath10 harmonic signals of heaters and thermometer . this gives high resolution and allows continuous output adjustments without additional noise . iv ) absolute accuracy is achieved via a variable - frequency - fixed - phase technique in which the measurement frequency is automatically adjusted during the measurements to account for the temperature variation of the sample heat capacity and the device thermal conductance . the performance of the calorimeter is illustrated by studying the heat capacity of a small au sample and the specific heat of a @xmath11 piece of superconducting pb in various magnetic fields .
in its most basic form , the scenario we consider can be described as follows . on one side we have a provider , bob , who controls an ordered classical database composed of @xmath0 memory cells . each cell of the data - base contains an @xmath1 bit string , so that the database consists of @xmath2 strings @xmath3 . on the other side , we have the person querying the database , alice , who wants to recover the string associated with a memory cell ( say the @xmath4-th one ) but at the same time does not want bob to know which cell she is interested in ( _ user privacy _ ) . in a purely classical setting the simplest strategy for alice consists in placing a large number of decoy queries , i.e. she `` hides '' her query among a large number @xmath5 of randomly selected queries . in this case , she will be able to get the information she is looking for , while limiting bob s intrusion in her privacy . [ in fact , the mutual information between alice s true query @xmath4 and bob s estimate of such value is upper bounded by @xmath6 . the drawbacks associated with such procedures are evident . first of all , the method does not allow alice to check whether bob is retaining information on her queries . moreover , to achieve a high level of privacy alice is forced to submit large amounts of fake queries , increasing the communication cost of the transition : in particular , absolute privacy is obtained only for @xmath7 , i.e. by asking bob to send _ all _ his database . this may not be acceptable if the database is huge or if it is an asset for bob ( _ data privacy _ ) . user and data privacy are apparently in conflict : the most straightforward way to obtain user privacy is for alice to have bob send her the entire database , leading to no data privacy whatsoever . conversely , techniques for guaranteeing the server s data privacy typically leave the user vulnerable @xcite . at the information theoretical level , this problem has been formalized as the symmetrically - private information retrieval ( spir ) @xcite generalizing the private information retrieval ( pir ) problem @xcite which deals with user privacy alone . spir is closely related to oblivious transfer @xcite , in which bob sends to alice @xmath2 bits , out of which alice can access exactly one which one , bob does nt know . no efficient solutions in terms of communication complexity @xcite are known for spir . indeed , even rephrasing them at a quantum level @xcite , the best known solution for the spir problem ( with a single database server ) employs @xmath8 qubits to be exchanged between the server and the user , and ensures data privacy only in the case of _ honest users _ ( i.e. users who do not want to compromise their chances of getting the information about the selected item in order to get more ) . better performance is obtained for the case of multiple non - mutually communicating servers @xcite ( although the user can not have any guarantee that the servers are not secretly cooperating to violate her privacy ) , while sub - linear communication complexity is possible under the some computational complexity assumption , e.g. @xcite . pir admits protocols that are more efficient in terms of communication complexity @xcite . the quantum private queries ( qpq ) protocol we have introduced in ref . @xcite is a cheat sensitive strategy @xcite which addresses both user and data privacy while allowing an exponential reduction in the communication and computational complexity with respect to the best ( quantum or classical ) single - server spir protocol proposed so far . specifically qpq provides a method to check whether or not bob is cheating and does not need the exchange of the whole database ( i.e. @xmath8 qubits ) : in its simplest form it only requires bob to transfer two database elements , identified by @xmath9 qubits , for each query . the qpq protocol is ideally composed by a preliminary signaling stage where the user and the database provider exchange some quantum messages ( specifically alice addresses bob receiving some feedback from him ) and by a subsequent _ retrieval&check _ stage where alice performs some simple quantum information processing on the received messages to recover the information she is interested in and to check bob s honesty . the qpq security relies on the fact that if bob tries to infer the query alice is looking for , she has a nonzero probability of discovering it . most importantly , one can verify that the more information bob gets on alice query , the higher is the probability that he will not pass alice s honesty test . in this paper we will derive analytical bounds for such a theoretical trade - off , and we analyze different variants of the qpq protocol . the main idea behind the protocol is the following . alice submits her request to bob using some quantum information carrier , so that she can either submit a plain query @xmath10 or a quantum superposition of different queries @xmath11 . alice randomly alternates superposed queries and non - superposed queries . thus , bob does not know whether the request he is receiving at any given time is a superposition of queries or not , so that he does not know which measurement will leave the information carrier unperturbed : he can not extract information without risking to introduce a disturbance that alice can detect . bob can , however , respond to alice s request without knowing which kind of query was submitted . his response will be either of the form @xmath12 or of the form @xmath13 , where the first ket is the register that alice had sent him , the second ket is a register that contains bob s answer ( @xmath14 being the answer to the @xmath15th query ) , and which may be entangled with the first . from these answers alice can both obtain the reply to her query and check that bob has not tried to breach her privacy . the main assumption we adopt is that , for each @xmath4 , there exists a _ unique _ answer string @xmath16 that can be independently checked by alice . [ this does not prevent different queries from having the same answer : indeed we do admit the possibility to have @xmath17 for @xmath18 . ] for example , alice may be asking bob the prime factors of one out of @xmath2 very large integer numbers ( say the rsa collection ) which she can not factorize by herself . the above requirements can be relaxed ( examples will be provided in sec . [ sec : soluzione ] ) , but they are useful as they permit a considerable simplification of the security proof . for the same reason , we will focus on the simplest version of the qpq protocol , where there exists a reference query @xmath19 ( dubbed _ rhetoric _ query ) which has a known standard answer @xmath20 . as discussed in ref . @xcite , this assumption is not fundamental , but it is very useful since it allows us to minimize the amount of exchanged signals in the protocol ( as a matter of fact alternative versions of the qpq protocol with higher security level can be devised which do not employ the rhetoric query ) . the paper is organized as follows . in sec . [ sec : notations ] we describe the rhetoric version of qpq in its basic form and introduce the notation . this is followed by the technical sec . [ sec : bob ] where we analyze in detail the most general transformations bob can perform on alice s queries . section [ sec : security ] contains the main result of the paper : here we introduce the trade - off between bob s information on alice s query and the success probability of him passing her honesty test . in sec . [ sec : ent ] we present some variations of the qpq protocol , one of which exploits entanglement as a resource to strengthen alice s privacy . finally in sec . [ sec : what ] we analyze what happens when relaxing some of the assumptions adopted in the security proof . in particular we show that the basic version of the qpq described here does not guarantee privacy if the queries have multiple answers , and we point out a possible solution in sec . [ sec : soluzione ] . th record of bob s database ( composed of @xmath0 records ) . she then prepares two @xmath21-qubit registers . the first contains the state @xmath22 , while the second contains the quantum superposition @xmath23 between her query and the rhetoric question `` @xmath19 '' , to which she knows the standard answer @xmath20 . she then sends , in random order ( i.e. randomly choosing either scenario * a * or scenario * b * ) , these two registers to bob , waiting for his first reply before sending the next register . bob uses each of the two registers to interrogate his database using a qram device , which records the reply to her queries in the two `` reply '' registers @xmath24 . at the end of their exchange , alice possesses the states @xmath25 and @xmath26 , where the @xmath16 is the content of the @xmath4th record in the database . by measuring the first she obtains the value of @xmath16 , with which she can check whether the superposition in the second state was preserved . if this is not the case , then she can be confident bob that bob has violated her privacy , and has tried to obtain information on what @xmath4 was . ] in the rhetoric version of the qpq protocol ( see fig . [ f : proto ] ) alice uses two quantum registers each time she needs to interrogate bob s database . the first register contains @xmath10 , the address of the database memory cell she is interested in ; the other register is prepared in a quantum superposition of the type @xmath27 , `` @xmath19 '' being the rhetoric query . alice then _ secretly _ and _ randomly _ chooses one of the two registers and sends it to bob . he returns the register alice has sent to him , together with an extra register in which the corresponding answer is encoded . in order to reply to alice s query without knowing whether it is the superposed query or not , bob needs to employ the quantum random access memory ( qram ) algorithm @xcite . after alice has received bob s first reply , she sends her second register and waits for bob s second reply . again bob returns her register together with an extra register which encodes his reply obtained through a qram application . if bob has followed the protocol accurately , without trying to extract information , alice now should possess a state which encodes the information she is looking for and an entangled state involving the rhetoric query , whose coherence can be tested to check bob honesty , i.e. the two states @xmath12 and @xmath28 . alice recovers the value of @xmath16 by measuring the second register in the first state , and then she uses this value to prepare a measurement to test whether the superposition has been retained in the second state ( `` honesty test '' ) . such a measurement is simply a projective measurement on the state @xmath28 . if this test fails , namely if she finds out that the state bob has sent her back is orthogonal to the one she is expecting , she can be confident that bob has cheated and has violated her privacy . if , instead , the test passes , she can not conclude anything . in fact , suppose that bob has measured the state and collapsed it to the form @xmath12 or to the form @xmath29 , it still has a probability 1/2 to pass alice s test of it being of the form @xmath28 . so alice s cheat test allows her to be confident that bob has cheated if the test fails , but she can never be completely confident that bob has not cheated if the test passes . we now introduce the notation which will be used . we define @xmath30 the source space which contains the addresses @xmath4 of the memory cells which compose bob s database , identifying with @xmath31 the address of the rhetoric query . for each @xmath4 we define @xmath16 to be the information associated with the @xmath4-th address . as mentioned in the introduction the @xmath16 are _ classical messages _ composed of @xmath1 bits , and they need not represent distinct messages ( i.e. we allow the possibility that @xmath32 for @xmath33 ) , but they are uniquely determined by the value of @xmath4 . in this context , bob s database is defined as the ordered set @xmath34 formed by the strings @xmath16 . we define @xmath35 the two quantum registers alice uses to submit her queries ; according to the protocol , she will first send @xmath36 , wait for bob s answer and then send @xmath37 . in this notation , for @xmath38 , the vector @xmath39 is the state of the @xmath40-th register which carries the address of the @xmath4-th database memory . for all @xmath41 we use the vector @xmath42 to represent the superposition of the @xmath4-th query and the rhetoric query , i.e. @xmath43 ( for @xmath31 we have @xmath44 ) . we define @xmath45 the registers on which bob writes the information to send back to alice . after having received @xmath36 from alice , bob encodes the necessary information on @xmath46 and sends back to her both @xmath36 and @xmath46 . analogously , after having received @xmath47 , he will encode information on @xmath48 and send her back both @xmath47 and @xmath48 . it is useful to also define the vectors @xmath49 ( as in eq . ( [ defq1 ] ) for @xmath31 we set @xmath50 ) . according to the protocol , the vectors @xmath51 or @xmath52 are the states that an honest bob should send back to alice when she is preparing @xmath53 into the states @xmath39 or @xmath42 , respectively . in fact , the states @xmath51 and @xmath52 are the result of the qram transformation when it is fed @xmath39 and @xmath42 , respectively . we also introduce an ancillary system @xmath54 to represent any auxiliary systems that bob may employ when performing his local transformation on the alice queries , plus ( possibly ) an external environment . let us use this notation to better formalize the qpq protocol described above . suppose then that alice wants to address the @xmath4-th entry of the database . the protocol goes as follows : alice randomly chooses between the two alternative scenarios * a * and * b * ( see fig . [ f : proto ] ) . in the scenario * a * , she prepares the qubits @xmath36 in @xmath55 and the qubits @xmath37 in @xmath56 . instead , in the scenario * b * she prepares the states @xmath57 and @xmath58 . this means that , in the scenario * a * , she first sends the plain query and then the superposed query . on the contrary , in the scenario * b * , she first sends the superposed query and then the plain query . consequently , the input state of the system @xmath59 is described by the vectors @xmath60 where the index @xmath61 refers to the selected scenario and @xmath62 is the fiducial initial state of the systems @xmath63 and @xmath54 ( it is independent on @xmath61 because bob does not know which scenario alice has chosen ) . 2 . now alice sends @xmath36 and waits until bob gives her back @xmath36 and @xmath46 . then , she sends @xmath37 and waits until she gets back @xmath37 and @xmath48 . 3 . _ honesty test _ : alice checks the states she has received . if she had selected scenario * a * , she performs a von neumann measurement to see if @xmath64 is in the state @xmath65 see eq . ( [ check ] ) . of course , this can be done in two steps : first she measures @xmath66 to learn @xmath16 and then she uses this value to prepare an appropriate measurement on @xmath67 . if the measurement fails , then alice can definitely conclude that bob was cheating , otherwise she can assume he was honest ( although she has no guarantee of it ) . if she had chosen scenario * b * , she proceeds analogously , using a von neumann measurement to check if @xmath64 is in the state @xmath68 . in the qpq protocol , alice s privacy relies essentially on the fact that bob is not allowed to operate jointly on @xmath36 and @xmath37 . this a fundamental constraint : without it , bob would be able to discover the index @xmath4 without alice knowing it . in fact , the subspaces @xmath69 spanned by the two vectors @xmath70 and @xmath71 ( associated to the two different scenarios * a * and * b * for the query @xmath4 ) are mutually orthogonal . thus , such vectors ( and then the corresponding queries ) could be easily distinguished by performing on @xmath72 a simple von neumann measurement defined by the projectors associated with the spaces @xmath73 . this is a measurement that would allow bob to recover alice s query without disturbing the input states of @xmath72 . to prevent this cheating strategy , the qpq protocol forces bob to address @xmath36 and @xmath37 separately ( i.e. he has to send the register @xmath36 back , before alice provides him the register @xmath37 ) . bob s action when he receives alice s first register can be described by a unitary operator @xmath74 which acts on the first register @xmath36 , on @xmath63 , and on @xmath54 ( and not on the second register @xmath37 which is still in alice s possession ) . analogously , bob s action when he receives the second register is described by the unitary operator @xmath75 which acts on @xmath37 , @xmath48 , and @xmath54 ( and not on @xmath36 and @xmath46 which are now in alice s possession ) . [ note that the above framework describes also the situation in which bob is employing non - unitary transformations ( i.e. cp - maps ) , since the space @xmath54 can be thought to contain also the naimark extension that transforms any cp - map into a unitary . ] the above transformations can not depend on the selected scenario @xmath61 ( as bob does not know which one , among @xmath76a and @xmath76b , has been selected by alice ) . therefore , within the @xmath61th scenario , the global state at the end of the protocol is described by the vectors @xmath77 with @xmath78 given in eq . ( [ equation1 ] ) . consider the transformation @xmath79 . in the scenario * a * for all @xmath4 we can write @xmath80 where @xmath81 stands for the separable state @xmath82 and where @xmath83 is a vector orthogonal to @xmath84 , i.e. @xmath85 with this choice , @xmath86 is the probability that the state ( [ eecc ] ) will be found in @xmath87 . in the scenario * b * , instead , for @xmath41 we can write @xmath88 as before @xmath89 and @xmath90 is a vector orthogonal to the `` check state '' @xmath91 alice is expecting , i.e. @xmath92 consequently @xmath93 is the probability that the state ( [ eecc1 ] ) will pass the test of being in @xmath94 . the state on the first line of eq . ( [ eecc1 ] ) can be expanded on a basis of which the state on the first line of eq . ( [ eecc ] ) is a component . therefore @xmath93 and @xmath86 must be related . the security analysis given in the following sections is based on the study of this relation . 1 analogous decompositions can be given for @xmath95 : in this case , however , it is useful to describe them not in terms of the input states , but in terms of the state of the system _ after it has passed the test on the subsystems @xmath66_. for @xmath41 , in the scenario * a * this gives : @xmath96 here @xmath97 is a vector orthogonal to @xmath98 of eq . ( [ check ] ) i.e. @xmath99 thus @xmath100 is the probability that the state ( [ eecc ] ) will pass the test of being in @xmath101 . notice that the vector @xmath102 in the first line of eq . ( [ eecc ] ) is the state of @xmath103 one obtains in the scenario * a * if , after the first round , the state @xmath66 passes the test of being @xmath104 see eq . ( [ eecc ] ) . in the scenario * b * , instead , we have @xmath105 where @xmath106 is a vector orthogonal to the state @xmath107 , i.e. @xmath108 and @xmath109 is the probability that the state ( [ eecc1 ] ) will be found in @xmath107 . 1 the case @xmath31 has to be treated separately : indeed , if alice sends this query then both @xmath36 and @xmath37 will be prepared into @xmath110 . in this case , it is then useful to define @xmath95 by considering its action on the vector @xmath111 defined as in eq . ( [ eecc ] ) , i.e. @xmath112 where again one has @xmath113 1from the above equations , it follows that for @xmath41 the final state ( [ finale ] ) , after bob has finished his manipulations , can be written as follows for scenario * a * @xmath114 where all the terms in the second and third line are orthogonal to @xmath115 . analogously , we have for scenario * b * @xmath116 where , again , the states in the last two lines are orthogonal to the state in the first . instead , for @xmath31 we have @xmath117 in this section we present an information - disturbance analysis of the qpq protocol . this will yield a trade - off which shows that , if bob tries to get some information on alice s queries , then she has a nonzero probability of detecting that he is cheating . the same analysis can be easily reproduced for more complicated versions of the protocol . for instance alice may hide her queries into superpositions of randomly selected queries . in this case , the derivation , although more involved , is a straightforward generalization of the one presented here . according to eq . ( [ finale ] ) , to measure bob s information gain , it is sufficient to study how the final state of the ancillary subsystem @xmath54 depends upon alice s query @xmath4 . exploiting the decompositions introduced in sec . [ sec : bob ] we can then show that one can force @xmath54 to keep no track of alice s query by bounding the success probabilities that bob will pass the qpq honesty test . specifically , indicating with @xmath118 the success probability associated with alice s query @xmath4 in the @xmath61-th scenario and defining @xmath119 the corresponding output state of @xmath54 , in sec . [ sec : ancilla ] we will prove the following theorem * theorem : * _ choose @xmath120 $ ] so that @xmath121 for all @xmath4 and @xmath61 . then there exists a state @xmath122 of @xmath54 and a positive constant @xmath123 such that the fidelities @xcite @xmath124 are bounded as follows , latexmath:[\ ] ] where the system @xmath276 is an ancillary system that alice does not hand over to bob . the protocol now follows the same procedure as the `` canonical '' qpq described previously , but employing the state @xmath277 in place of the state @xmath158 . of course , alice s honesty test must be appropriately modified , as she has to test whether bob s actions have destroyed the entanglement between the ancillary system @xmath276 and the @xmath278 register . the main difference with the canonical qpq is that here half of the times bob has only access to a part of an entangled state : he is even more limited in re - preparing the states for alice than in the canonical qpq . it is easy to see that the security proof given in the previous sections can be straightforwardly extended to this version of the protocol , and that the security bounds derived above still apply : indeed they can be made even more stringent as bob has only a limited capacity in his transformations on alice s queries , since he does not have access to the ancillary space @xmath276 . in the situations in which the information carriers employed in the queries can be put in a superposition of traveling in different directions @xcite , this version of the protocol can easily be reduced to the canonical qpq by simply supposing that alice is in the possession of the database element @xmath31 corresponding to the rhetoric question , while , obviously , bob is in possession of all the remaining database elements . in deriving the qpq protocol it is assumed that for each query @xmath279 there is a _ unique _ possible answer @xmath16 ( notice however that two distinct queries can have the same answer i.e. @xmath16 can coincides with @xmath280 ) . one way to enforce this condition in a realistic scenario is to admit the possibility that alice can independently verify the answer that bob is sending to her is difficult to compute for alice , but once a solution is given to her , she can easily check if it is correct . ] . in this section we will show that if this is not the case then the basic structure of qpq does not prevent bob to cheat without being discovered by alice . in sec . [ sec : soluzione ] we will discuss how one can overcome these limitations , at least temporarily , by allowing alice ( or third parties that collaborate with her ) to reiterate her query at random times . here we drop the above hypothesis and give two examples of successful cheating strategies that allow bob to spy on alice s query , and still pass the honesty test with probability @xmath236 . let us start by considering the case of a database with @xmath281 possible entries in which both the query @xmath282 and the query @xmath283 admit two distinguishable answers . in particular let @xmath284 , @xmath285 be the answers for @xmath282 and @xmath286 , @xmath287 those for @xmath283 . now , suppose that the unitary @xmath288 of eq . ( [ finale ] ) that bob applies to @xmath289 performs the following mapping @xmath290 where @xmath291 is the answer to the rhetoric query and where @xmath292 with @xmath293 and @xmath294 being orthonormal states of bob s space @xmath54 . analogously define @xmath295 as the unitary operator which performs the following transformation @xmath296 of @xmath54 and @xmath297 according to the above assumptions , if alice s query is the rhetoric one ( i.e @xmath31 ) the final state ( [ finale ] ) of the qpq protocol is @xmath298 as output state . for @xmath299 , instead , we have two possibilities . in the scenario @xmath76a the final state will be latexmath:[\[\begin{aligned } & \tfrac { & + \tfrac{\|j\rangle_{q_2 } \| a_j^{(- ) } \rangle_{r_2 } + \|0\rangle_{q_2 } while in the scenario @xmath76b it will be @xmath301 alice will receive the answer @xmath302 half of the times and the answer @xmath303 in the other half of the times , while bob will always pass the honesty test . moreover in the case in which alice receives the answer @xmath302 , bob will get the state @xmath304 while in the case in which alice receives the answer @xmath303 bob will get the state @xmath305 . in average the state @xmath54 is @xmath306 . in conclusion , using @xmath79 and @xmath95 as in the previous paragraphs , bob will always pass the honesty test . furthermore the output state of @xmath54 he gets at the end of the protocol will be partially correlated with the query @xmath4 as follows : @xmath307 therefore by performing a simple von neumann measurement on @xmath54 , bob will be able to extract some information on @xmath4 , without alice having any chance of detecting it . notice that , in the example presented here , bob s info is limited by the partial overlap between the states @xmath308 , @xmath309 and @xmath310 . however , this is not a fundamental limitation as one can construct more complex examples ( e.g. databases with more than two possible answers for a single query ) for which the amount of info that bob acquires on @xmath4 can be arbitrarily high . it is also important to stress that the above example can be used also to show that bob will be able to cheat also in the case in which alice adopts qpq strategies more sophisticated then the simple rhetoric version discussed in this paper ( e.g. instead of sending superpositions of the form @xmath311 with @xmath271 and @xmath272 arbitrary amplitudes that only she knows ) . here we analyze how multiple valid answers may affect the performance of the non - rhetoric version of the qpq protocol ( i.e. where alice is not using the rhetoric question @xmath31 ) . we give an example of a successful cheating strategy for a database with @xmath281 queries . for the sake of simplicity , we will assume that @xmath312 have single answers @xmath20 and @xmath313 respectively , but that @xmath283 is associated with two distinguishable answers @xmath314 . as an example of a non - rhetoric qpq protocol we consider the case in which alice , to get the information associated with the @xmath4 query , chooses another query ( say the @xmath315-th one ) and sends sequentially , in random order , states of the form @xmath316 , @xmath10 and @xmath317 ( @xmath271 and @xmath272 being amplitudes that only she knows ) . as in the case of the rhetoric version of the protocol , bob s action can be described by unitaries . in this case they are @xmath79,@xmath95 and @xmath318 . notice that the first acts on @xmath289 the second on @xmath319 and the third on @xmath320 , with obvious choice of the notation for the subspaces involved . for our present purpose , it is sufficient to assume that for @xmath321 , @xmath322 acts non - trivially only on @xmath323 ( this is a particular instance of the general case ) . we can also assume that @xmath79,@xmath95 and @xmath318 are identical . we then define such operators according to the following rules : latexmath:[\[\begin{aligned } u^{(k)}_{q_kr_kb } ( \|j\rangle_{q_k } \|0\rangle_{r_k}\|0\rangle_b ) & = & u^{(k)}_{q_kr_kb } ( \|j\rangle_{q_k } \|0\rangle_{r_k}\|2\rangle_b ) & = & if @xmath312 while , for @xmath283 , @xmath325 are defined in eq . ( [ defff ] ) . if initial state of the @xmath54 is @xmath326 , one can easily verify that bob will always pass alice s honesty test ( no matter which superposition @xmath327 she is using ) and that he can recover part of the information associated with the query . in this simple example , for instance , he has a not null probability to identify the query @xmath283 . as before , this counterexample can be easily generalized and improved . the case in which different answers may correspond to the same query is , of course , quite relevant , so that it is natural to ask if the qpq protocol can be modified to apply also to this situation . in this section we give some methods that allow alice to foil the cheating strategies described in the previous section temporarily , for as long as bob is expecting further queries . in the case in which alice can independently check how many different replies correspond to the each query ( and which are they ) , then there is a simple solution that prevents bob from cheating : we must require bob to provide all possible replies in a pre - established order ( e.g. alphabetically ) when he is presented the @xmath4th query . in this way , each query has again a unique _ composite _ answer ( composed by the ordered succession of all the possible answers ) , so that we are reduced to the canonical qpq protocol , and bob is prevented from cheating . if , however , alice can not independently establish the number of different replies to each query , then a different strategy is necessary . [ note that the security proofs given in sects . [ sec : bob ] and [ sec : security ] cease to apply to this version of the protocol , although conceivably they may be extended to cover also this situation . ] first of all , we must require that each of the possible replies to the @xmath4th query is uniquely indexed by bob . this means that there should be a unique answer to the question `` what is the @xmath40th possible reply to the @xmath4th query ? '' of course , this by itself is insufficient to guarantee that bob can not employ the cheating strategies of the previous section , as alice can not independently check the uniqueness of bob s indexing ( since she does not know all the possible answers to the @xmath4th query ) . however , she can check whether bob will always answer in the same way to repeated queries . from eqs . ( [ llll ] ) and ( [ llla ] ) , it follows that , as soon as bob measures his system @xmath54 , he might gain information on the value of @xmath4 , but at the same time he loses information on which ( among all the possible answers to the @xmath4th query ) he had presented to alice . if he wants to be sure that he keeps on providing always the same answer to repeated queries on alice s part , he must preserve his system @xmath54 without trying to extract information from it . he can measure the system @xmath54 only when he is confident that alice will not be asking him the same query anymore . in a multi - party scenario , we can also think of a situation where multiple cooperating parties ask bob the same queries and compare the replies they receive from him . if they find that his answers when he is asked the @xmath40th reply to the @xmath4th query to do not match , then they can conclude that he has been cheating : he has not assigned a unique index to all the possible replies to the @xmath4th query , and he has taken advantage of the cheating strategies detailed in the previous section . bob is thus placed in the awkward situation of possessing information on alice s query in the system @xmath54 entirely in his possession , but of being prevented from accessing such information . this is a temporary solution , since , as soon as bob is certain that he will not be asked the @xmath4th query anymore , he can measure the system @xmath54 and extract the information stored on it . he is kept honest only as long as he is in business ( and , of course , he is in business only as long as he is honest ) . in conclusion , we have given a security proof of the qpq protocol introduced in @xcite . it is based on quantitative information - disturbance tradeoffs which place an upper bound on the information bob can retain on alice s query in terms of the disturbance he is producing on the states that he is handing back to her ( see sects . [ sec : bob ] and [ sec : security ] ) . a nonzero information retained by bob implies a nonzero disturbance on alice s states , which she can detect with a simple measurement ( the `` honesty test '' ) . if the honesty test fails , she can conclude that bob has certainly cheated . if , on the other hand , the test passes , she can tentatively conclude that bob has not cheated ( although she can not be certain of it ) . in addition , we have given some variants of the protocol to further increase alice s security , i.e. to reduce bob s probability of evading detection when cheating . these variants either exploit secret parameters , or exploit entanglement with an ancillary system alice retains in her possession ( see sect . [ sec : ent ] ) . finally , we have seen that bob can successfully cheat without being detected if we drop the assumption ( which is at the basis of the qpq protocol ) that to each query there can be associated only a single answer @xmath16 ( see sect . [ sec : what ] ) . in fact , if we assume that there exist two ( or more ) different replies @xmath328 to the query @xmath4 , then bob can find out the value of @xmath4 , evading detection by alice with certainty . we discussed some strategies that allow alice to protect herself also in this situation , at least as long as bob can expect further queries from her or from other parties who may cooperate with her ( see sect . [ sec : soluzione ] ) . we acknowledge discussions with e. kushilevitz , s. micali , f. sciarrino , and m. sudan . v. g. acknowledges support from the quantum information research program of centro di ricerca ennio de giorgi of sns . s. l. acknowledges fruitful discussions with s. brin and l. page . 100 y. gertner , y. ishai , e. kushilevitz , and t. malkin , journal of computer systems sciences , * 60 * 592 ( 2000 ) . b. chor , o. goldreich , e. kushilevitz , and m. sudan , journal of the acm , * 45 * , 965 ( 1998 ) ; c. cachin , s. micali , and m. stadler , in advances in cryptology - eurocrypt99 ( 1999 ) ; c. gentry and z. ramzan , in proc . 32nd icalp , 803 - 815 , ( 2005 ) ; s. yekhanin , technical report eccc tr06 - 127 ( 2006 ) . e. kushilevitz and r. ostrovsky in proc . 38th ieee symposium focs97 , 364 ( 1997 ) . s. wiesner , acm sigact news , * 15(1 ) , 78 - 88 , winter - spring ( 1983 ) ; m. o. rabin , ` how to exchange secrets with oblivious transfer , ' technical report tr-81 , harvard aiken computational laboratory ( 1981 ) ; a. jakoby , m. liskiewicz , a. madry , arxiv : quant - ph/0605150v1 . a. ambainis , in _ proceedings of the 24ht icalp _ , _ lecture notes in computer science _ , 1256 * 401 ( 1997 ) . i. kerenidis and r. de wolf , arxiv : quant - ph/0208062 ; i. kerenidis and r. de wolf , arxiv : quant - ph/0307076 . v. giovannetti , s. lloyd , and l. maccone , phys . * 100 * , 230502 ( 2008 ) , eprint arxiv:0708.2992 . l. hardy , a. kent , phys . rev . lett . * 92 * , 157901 ( 2004 ) . m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _ ( cambridge university press , cambridge , 2000 ) . v. giovannetti , s. lloyd , and l. maccone , _ phys . _ vol . * 100 * , pag 160501 ( 2008 ) . d. bures , _ trans . am . * 135 * , 199 ( 1969 ) ; a. uhlmann , _ rep . math . phys . _ * 9 * , 273 ( 1979 ) . t. cover and j. thomas , _ elements of information theory _ , ( john wiley & sons , new york , 1991 ) . v. giovannetti and a. s. holevo , _ quantum information processing _ * 7 * , 55 ( 2008 ) . a. s. holevo _ probabilistic and statistical aspects of quantum theory _ ( north holland , amsterdam , 1982 ) . r. f. streater , _ statistical dynamics _ ( imperial college press , england , 1995 ) . v. giovannetti , s. lloyd , l. maccone , f. sciarrino , _ experimental quantum private queries _ , unpublished ( 2008 ) .	we present a security analysis of the recently introduced quantum private query ( qpq ) protocol . it is a cheat sensitive quantum protocol to perform a private search on a classical database . it allows a user to retrieve an item from the database without revealing which item was retrieved , and at the same time it ensures data privacy of the database ( the information that the user can retrieve in a query is bounded and does not depend on the size of the database ) . the security analysis is based on information - disturbance tradeoffs which show that whenever the provider tries to obtain information on the query , the query ( encoded into a quantum system ) is disturbed so that the person querying the database can detect the privacy violation .
gamma - ray bursts ( grbs ) are the most powerful explosions in the universe since the big bang . their cosmological origins are identified by the redshift measurements of their exploded remnants , usually named `` grb afterglows '' , or their host galaxies . grbs are common regarded as jetted phenomena , supported by the observational evidence that an achromatic break appears in the afterglow light curve , which declines more steeply than in the spherical model largely due to the edge effect and the laterally spreading effect ( rhoads 1999 ; sari et al . 1999 ) or the off - axis viewing effect ( rossi , lazzati & rees 2002 ; zhang & mszros 2002a ; berger et al . 2003 ) . with the advantages of huge energy release for the prompt emission and immunity to dust extinction for the @xmath0-ray photons , grbs are widely believed to be detectable out to a very high redshift of @xmath15 ( lamb & reichart 2000 ; ciardi & loeb 2000 ; bromm & loeb 2002 ; gou et al . 2004 ) . similar to the phillips relation in type ia supernovae ( sne ia ; phillips 1993 ) , a tight relation in grbs , linking a couple of energetic and observational properties , could make grbs new standard candles . amati et al . ( 2002 ) reported a power - law relation between the isotropic - equivalent @xmath0-ray energy @xmath1 of a grb and its rest - frame @xmath3 peak energy @xmath2 . unfortunately , the large scatter around this relation stymies its cosmological purpose . however , after joining the burst s half - opening angle @xmath16 , ghirlanda et al . ( 2004a ) found that a tight relation is instead between the collimation - corrected energy @xmath17 and @xmath2 , which reads @xmath18 ( @xmath19 is the power ) . recently , liang & zhang ( 2005 ) reported a multi - variable relation between @xmath1 , @xmath2 , and the local achromatic break @xmath4 of the burst s afterglow light curve , i.e. , @xmath20 ( @xmath21 and @xmath22 are the powers ) . applying these two relations to different observed grb samples , meaningful cosmological constraints have been performed in a series of works ( e.g. dai , liang & xu 2004 ; ghirlanda et al . 2004b ; firmani et al . 2005 ; xu , dai & liang 2005 ; ghisellini et al . 2005 ; qin et al . 2005 ; mortsell & sollerman 2005 ; liang & zhang 2005 ; bertolami & silva 2005 ; lamb et al . 2005 for a grb mission plan to investigate the properties of dark energy ) . in this paper we investigate the well - measured spectral and energetic properties of the up - to - date 20 grbs in their cosmological frames and report a new relation between @xmath23 ( @xmath24 ) and @xmath2 , which is @xmath25 ( where @xmath26 is the power ) . more importantly , inspiring constraints on cosmological parameters can be achieved if this newly found relation is indeed an intrinsic one for the observed grb sample . the `` bolometric '' isotropic - equivalent energy of a grb is given by @xmath27 where @xmath28 is the fluence in the observed bandpass and @xmath29 is a multiplicative correction relating the observed bandpass with a standard rest - frame bandpass ( 1-@xmath30 kev ) ( bloom , frail & sari 2001 ) , with its fractional uncertainty @xmath31 thus , the collimation - related energy @xmath23 ( see @xmath32 of this work ) is @xmath33 where @xmath34 is the observed break time of the afterglow light curve in the optical band , with its fractional uncertainty @xmath35 to calculate @xmath29-correction of a burst , one is required to know its spectral parameters fitted by the band function , i.e. , the low - energy spectral index @xmath36 , the high - energy spectral index @xmath37 , and the observed peak energy @xmath38 ( band et al . 1993 ) . in this paper , we considered those grbs with known redshifts , fluences , spectral parameters , and break times , and thus collected a sample of 20 bursts shown in table 1 . the criterions of our data selection are : ( 1 ) the fluence and spectral parameters of a burst are chosen from the same original literature as possible . if different fluences are reported in that literature , we choose the measurement in the widest energy band . if this criterion is unsatisfied , we choose the fluence measured in the widest energy band available in other literature ; ( 2 ) the observed break time for each burst is taken from its data in the optical band ; ( 3 ) when the fractional uncertainties of @xmath39 , @xmath28 and @xmath29 are not reported in the literature , they are taken to be @xmath40 , @xmath41 and @xmath42 , respectively ; ( 4 ) to be more reliable , a lower limit of the fractional uncertainty of @xmath34 is set to be @xmath41 . in this section , we investigate the @xmath43 and @xmath44 relations for a friedmann - roberston - walker cosmology with mass density @xmath45 and vacuum energy density @xmath46 . therefore , the theoretical luminosity distance in equations ( 1 ) and ( 3 ) is given by @xmath47^{-1/2}\},\end{aligned}\ ] ] where @xmath48 , @xmath49 is the present hubble constant , and `` sinn '' is @xmath50 for @xmath51 and @xmath52 for @xmath53 . for @xmath54 , equation ( 5 ) degenerates to be @xmath55 times the integral ( carroll et al . according to the conclusions of the _ wilkinson microwave anisotropy probe _ ( wmap ) observations , we here choose @xmath7 , @xmath8 , and @xmath56 ( e.g. spergel et al . the dashed and solid lines in figure 1 respectively represent the best - fit powerlaws for the @xmath43 and @xmath44 relations ( weighting for the uncertainties on both coordinates , see press et al . we find @xmath57 with a reduced @xmath58 , and @xmath59 with a reduced @xmath60 . as can be seen , although the power of the @xmath43 relation in this work is roughly consistent with that in amati et al . ( 2002 ) derived from 9 _ bepposax _ bursts , i.e. , @xmath61 ( @xmath62 ) , the dispersion around this relation increases seriously . instead , the scatter of the @xmath44 relation is rather small . one may notice that it is the @xmath44 relation rather than the amati relation whose power more converges at @xmath63 . based on the above statistical findings , we consider the @xmath44 relation for the observed grb sample to constrain cosmological parameters . the @xmath44 relation can be given by @xmath64 where the dimensionless parameters @xmath65 and @xmath26 are assumed to have no covariance . combining equations ( 3 ) and ( 8) , we derive the apparent luminosity distance as @xmath66 with its fractional uncertainty being @xmath67 where all the quantities are assumed to be independent of each other and their uncertainties follow gaussian distributions . the distance modulus is calculated by @xmath68 , and its uncertainty is computed through @xmath69 . although the @xmath44 relation is similar to the phillips relation in sne ia , the methods by which to constrain cosmological parameters are different ( riess et al . 1998 ; firmani et al . 2005 , xu et al . 2005 ) . for sne ia , a phillips - like relation is known to those relatively high-@xmath70 objects after it has been calibrated by nearby well - observed sne ia , because the theoretical luminosity distance is irrelevant with cosmological parameters , i.e. @xmath71 , when @xmath72 . while for grbs , one wo nt know that sole set of @xmath73- parameters in the @xmath44 relation until a low-@xmath70 grb sample is established . therefore , the @xmath74 statistic for grbs is @xmath75 } ^2,\ ] ] where @xmath76 denotes a certain cosmology , and @xmath77 is taken as 0.71 in this work . we carry out a bayesian approach to obtain grbs constraints on cosmological parameters . for clarity , we divide it into three stages with the detailed steps as follows : * stage i * \(1 ) fix @xmath78 , ( 2 ) calculate @xmath79 and @xmath23 for each burst for that cosmology , ( 3 ) best fit the @xmath44 relation to yield @xmath80 and @xmath81 , ( 4 ) apply the best - fit relation parametrized by @xmath80 and @xmath81 to the observed sample , and thus derive @xmath82 and @xmath83 for each burst for that cosmology , ( 5 ) repeat steps 1@xmath844 from @xmath85 to @xmath86 to obtain @xmath79 , @xmath82 and @xmath87 for each burst for each cosmology ; * stage ii * \(6 ) re - fix @xmath88 , ( 7 ) calculate @xmath89 by comparing @xmath90 with @xmath91 , @xmath92 , and then convert it to a conditional probability , i.e. , probability for @xmath88 which is contributed by the relation calibrated for @xmath93 , by the formula of @xmath94 $ ] , ( 8) repeat step 7 from @xmath85 to @xmath95 to obtain an iterative probability for cosmology @xmath88 by @xmath96\times p^{\rm ini}(\omega _ i ) } $ ] ( here the initial probability for each cosmology is regarded as equal , i.e. , @xmath97 ) , ( 9 ) repeat steps 6@xmath848 from @xmath98 to @xmath99 to obtain an iterative probability @xmath100 for each cosmology ; * stage iii * the iterative probability @xmath100 for each cosmology derived on step 9 is no longer equal to its initial probability @xmath101 , but it has not reached the final / converged probability @xmath102 . so the following process is to ( 10 ) assign @xmath100 on step 9 to @xmath101 on step 8 , then repeat steps 8@xmath849 , and thus reach another set of iterative probabilities for each cosmology , ( 11 ) run the above iteration cycle again and again until the probability for each cosmology converges , i.e. , @xmath103 after tens of cycles . in this method , to calculate the probability for a favored cosmology , we consider contributions of all the possible @xmath44 relations associated with their weights . the conditional probability @xmath104 denotes the contribution of some certain relation , and @xmath105 weights the likelihood of this relation for its corresponding cosmology . therefore , this bayesian approach can be formulized by @xmath106 shown in figure 2 are the constraints for @xmath107 from 20 observed grbs ( solid contours ) , using the likelihood procedure presented here . the grb dataset is consistent with @xmath13 and @xmath14 , yielding a @xmath12 . for a flat universe , we measure @xmath9 at the @xmath108 confidence level ( c.l . ) . the best - fit point is @xmath109 and @xmath110 ( red cross ) . shape of the elliptic confidence contours trends to well constrain @xmath45 and provide evidence for cosmic acceleration with an enlarged grb sample . for comparison , we also plot the constraints from 157 gold sne ia in riess et al . ( 2004 ) ( dashed contours ) . closed @xmath111 ranges derived from these two datasets at the same c.l.s have conveyed the advantages of high-@xmath70 distance indicators in constraining cosmological parameters as well as other cosmological issues . additionally , grbs are complementary to sne ia on the cosmological aspect . we see from the figure that a combination of sne and grbs wound make the cosmic model of @xmath13 and @xmath14 more favored , thus better in agreement with the result form wmap observations ( e.g. bennett et al . grbs are hopeful to be new standard candles . we report a new relation between grb rest - frame energetics and spectra , which is @xmath112 ( @xmath6 ) for @xmath7 , @xmath8 and @xmath56 . considering this power - law relation for the 20 observed grbs , we find a constraint on the mass density @xmath9 ( @xmath10 ) for a flat universe together with a @xmath12 for @xmath13 and @xmath14 . as previously discussed , the ghirlanda relation reads @xmath113 ( @xmath19 is the power ) under the framework of uniform jet model . within this model , one can calculate the half - opening angle of a grb jet by @xmath114 , where @xmath115 is the uniform circumburst medium density , and @xmath116 denotes the conversion efficiency of the initial ejecta s kinetic energy to @xmath0-ray energy release ( rhoads 1999 ; sari et al . thanks to strong collimation for grbs , i.e. @xmath117 , the ghirlanda relation becomes @xmath118 . in ghirlanda et al . ( 2004a ) , the term of @xmath119 was assumed to be highly clustered for the whole sample of 15 bursts presented there , thus the ghirlanda relation turns out to be @xmath120 . so if @xmath121 ( @xmath122 ) , there will be @xmath123 ( @xmath124 ) . in this sense , the @xmath125 and @xmath44 relations are consistent with each other . however , @xmath116 should be different from burst to burst ( e.g. @xmath126 ) , and @xmath115 is variable for a burst in the wind environment and may expand in several orders ( see friedman & bloom 2004 and references therein ) . additionally , these two observables and their uncertainties are very difficult to be reliably estimated . therefore , if the ghirlanda relation is used to measure cosmology , the resulting constraints would depend on different input assumptions of @xmath116 and @xmath115 . for contrast , such difficulties have been circumvented when applying the @xmath44 relation . moreover , liang & zhang ( 2005 ) reported a generalized relation between grb rest - frame spectra and energetics could be written as @xmath127 for a flat universe of @xmath128 , using a multiple variable regression method for the 15 bursts presented there . their conclusion is consistent with our statistical result ( see @xmath129 in this work ) . more loose constraints , however , were performed when the uncertainties of the fitted parameters in this relation were included into the error of the apparent distance modulus ( see fig 11 in liang & zhang 2005 ) . among the discussed three relations , the @xmath44 relation is the simplest and the latter two have the advantage of making the relations explicitly model - independent and eliminating the needs to marginalize over the unknown @xmath116 and @xmath115 . so what is the underlying theoretical basis for the @xmath44 relation ? at the present stage , plausible explanations mainly include : the standard synchrotron mechanisms in relativistic shocks ( e.g. zhang & mszros 2002b ; dai & lu 2002 ) , the high - energy emission form off - axis relativistic jets ( e.g. yamazaki et al . 2004 ; eichler & levinson 2004 ; levinson & eichler 2005 ) , and the dissipative photosphere model producing a relativistic outflow ( e.g. rees & mszros 2005 ) . the scaling relation depends on the details of each model , and it resembles the observational result under certain simplification . although models are different , the relation they support has made grbs towards more and more standardized candles . xd is thankful to z. g. dai , e. w. liang and b. zhang for valuable discussions and d. band in explaining the observations of grb050525a . this work is supported by the national natural science foundation of china ( grants 10233010 and 10221001 ) , and the ministry of science and technology of china ( nkbrsf g19990754 ) . amati , l. et al . 2002 , a&a , 390 , 81 amati , l. 2003 , chjaa , 3 , 455 andersen , m. i. et al . 2003 , gcn , 1993 band , d. , matteson , j. , ford , l. , et al . 1993 , apj , 413 , 281 barth , a. j. et al . 2003 , apj , 584 , l47 barraud , c. et al . 2003 , a&a , 400 , 1021 bennett , c. l. et al . 2003 , apjs , 148 , 1 berger , e. et al . 2002 , apj , 581 , 981 berger , e. et al . 2003 , nature , 426 , 154 bertolami o. & silva p. t. 2005 , preprint , astro - ph/0507192 bj " ornsson , g. et al . 2001 , apj , 552 , l121 bloom , j. s. , frail , d. a. , & sari , r. 2001 , aj , 121 , 2879 bloom , j. s. , frail , d. a. , & kulkarni , s. r. 2003 , apj , 594 , 674 blustin , a. j. , band , d. , barthelmy , s. et al . 2005 , preprint , astro - ph/0507515 bromm , v. , & loeb , a. 2002 , apj , 575 , 111 butler n. , vanderspek , r. , marshall , h. l. et al . 2004 , gcn , 2808 , see also http://space.mit.edu/hete/bursts/grb041006 carroll , s. m. , press , w. h. , & turner , e. l. 1992 , ara&a , 30 , 499 ciardi , b. , & loeb , a. 2000 , apj , 540 , 687 crew , g. b. et al . 2003 , apj , 599 , 387 dai , z. g. , & lu , t. 2002 , apj , 580 , 1013 dai , z. g. , liang , e. w. & xu , d. 2004 , apj , 612 , l101 djorgovski , s. g. et al . 2001 , apj , 562 , 654 eichler , d. , & levinson , a. 2004 , apj , 614 , l13 firmani , c. et al . 2005 , mnras , 360 , 1 frail , d. a. et al . 2003 , apj , 590 , 992 friedman , a. s. & bloom , j. s. 2005 , apj , 627 , 1 ghirlanda , g. , ghisellini , g. , & lazzati , d. 2004a , apj , 616 , 331 ghirlanda , g. , ghisellini , g. , lazzati , d. & firmani , c. 2004b , apj , 613 , l13 ghisellini g. , ghirlanda g. , firmani c. , avila - reese v. , 2005 , preprint , astro - ph/0504306 , review at the 4th workshop gamma - ray bursts in the afterglow era godet , o. et al . 2005 , gcn , 3222 gou , l. j. , mszros , p. , abel , t. & zhang , b. 2004 , apj , 604 , 508 halpern , j. p. , et al . 2000 , apj , 543 , 697 holland , et al . 2003 , aj , 125 , 2291 holland , et al . 2004 , aj , 128 , 1955 jakobsson , p. et al . 2003 , a&a , 408 , 941 jakobsson , p. et al . 2004 , a&a , 427 , 785 jimenez , r. , band , d. l. , & piran , t. , 2001 , apj , 561 , 171 klose , s. et al . 2004 , aj , 128 , 1942 kulkarni , s. r. et al . 1999 , nature , 398 , 389 lamb , d. q. , & reichart , d. e. 2000 , apj , 536 , 1 lamb d. q. , ricker g. r. , lazzati d. , et al . , 2005 , astro - ph/0507362 liang , e. w. & zhang , b. 2005 , apj in press ( astro - ph/0504404 ) levinson , a. & eichler , d. 2005 , apjl in press ( astro - ph/0504125 ) masetti , n. et al . 2000 , a&a , 354 , 473 mortsell e. & sollerman j. , 2005 , jcap , 0506 , 009 phillips , m. m. 1993 , apj , 413 , l105 press , w. h. et al 1999 , _ numerical recipes in fortran _ , cambridge university press price , p. a. et al . 2003a , apj , 589 , 838 price , p. a. et al . 2003b , nature , 423 , 844 qin y. p. , zhang b. b. , dong y. m. et al . , 2005 , preprint , astro - ph/0502373 rees , m. j. & mszros , p. 2005 , apj in press ( astro - ph/0412702 ) rhoads , j. e. 1999 , apj , 525 , 737 riess , a. g. et al . 1998 , aj , 116 , 1009 riess , a. g. et al . 2004 , apj , 607 , 665 rossi , e. , lazzati , d. & rees , m. j. 2002 , mnras , 332 , 945 sakamoto , t. et al . 2004 , preprint ( astro - ph/0409128 ) sakamoto , t. , ricker , g. , atteia , j - l . 2005 , gcn , 3189 , see also http://space.mit.edu/hete/bursts/grb050408 sari , r. , piran , t. , & halpern , j. p. 1999 , apj , 519 , l17 spergel , d. n. et al . 2003 , apjs , 148 , 175 stanek , k. z. et al . 1999 , apj , 522 , l39 stanek , k. z. et al . 2005 , apj , 626 , l5 vanderspek , r. et al . 2004 , aj , 617 , 1251 xu , d. , dai , z. g. & liang , e. w. 2005 , apj in press ( astro - ph/0501458 ) yamazaki , r. , ioka , k. , & nakamura , t. 2004 , apj , 606 , l33 zhang , b. & mszros , p. 2002a , apj , 571 , 876 zhang , b. & mszros , p. 2002b , apj , 581 , 1236 plane . circles represent @xmath23 vs. @xmath2 for 20 grbs , while squares denote @xmath1 vs. @xmath2 . solid and dashed lines are the best fits of the @xmath44 and @xmath43 relations , using @xmath7 , @xmath8 , and @xmath56.[fig1],scaledwidth=80.0% ] plane from 20 observed grbs ( solid contours ) , and from 157 gold sne ia ( dashed contours ) . red and black crosses mark the best fits of the two samples , while black dot denotes the concordance model of @xmath130.[fig2],scaledwidth=70.0% ]	we investigate the well - measured spectral and energetic properties of 20 gamma - ray bursts ( grbs ) in their cosmological rest frames . we find a tight relation between the isotropic - equivalent @xmath0-ray energy @xmath1 , the local peak energy @xmath2 of the @xmath3 spectrum , and the local break time @xmath4 of the grb afterglow light curve , which reads @xmath5 ( @xmath6 ; @xmath7 , @xmath8 ) . such a power - law relation can be understood via the high - energy radiation processes for the grb prompt emission accompanying the beaming effects . we then consider this relation as an intrinsic one for the observed grb sample , and obtain a constraint on the mass density @xmath9 ( @xmath10 ) for a flat @xmath11cdm universe , and a @xmath12 for @xmath13 and @xmath14 . ongoing grb observations in the _ swift _ era are expected to confirm this relation and make its cosmological utility progress much .
understanding the origin of ferromagnetism was one of the success stories of applying quantum mechanics to solid - state systems . classically , magnetic moments interact via the dipole - dipole interaction , which is much too weak to explain magnetic order at as high a temperature as is observed in , e.g. , iron or nickel.@xcite the explanation of this conundrum was found to be the exchange interaction mechanism , which leads to a spin - spin interaction that is governed by the coulomb interaction via the pauli principle . this was first understood in the context of atomic and molecular physics in the 1920s , and applied to solid - state physics in the 1950s.@xcite somewhat ironically , a straightforward application of the exchange interaction concept leads to a spin - spin interaction that is too strong , as the relevant energy scale is the atomic scale , or roughly @xmath0 . many - body and band - structure effects renormalize this scale and bring it down to the observed ferromagnetic scale of rougly @xmath1 or lower.@xcite this is still much larger than the dipole - dipole scale , and the latter is often neglected in the discussion of ferromagnets . when it is considered , e.g. , for its influence on the critical behavior,@xcite it is usually added phenomenologically to models that describe the exchange interaction . another spin - spin interaction that has been of interest lately is the dzyaloshinsky - moriya ( dm ) interaction.@xcite it results ( in systems with suitable lattice structures ) from the spin - orbit interaction , has been derived from microscopic models , and is believed to be responsible for the helical magnetic order observed in mnsi and fege.@xcite rough estimates show that the dm interaction and the dipole - dipole interaction are of about the same strength , and should thus be considered together.@xcite furthermore , in weak ferromagnets , which order only at low temperatures , all three interactions can be comparable in strength , which can make the dipole - dipole and dm interactions crucial . in this paper we provide a comprehensive derivation of all of these effects within one unified framework , namely , a field - theoretic description of electrons and photons . starting with finite - temperature quantum electrodynamics ( qed ) coupled to a field - theoretic description of finite - density quasi - relativistic electrons we show that the exchange , dipole - dipole , and dm interactions all appear naturally upon integrating out the photons . the exchange and dm interactions arise from integrating out the scalar part of the electromagnetic potential ; the dipole - dipole interaction , from integrating out the vector potential . furthermore , the dm and dipole - dipole interactions are indeed of the same order in the relativistic corrections to the schrdinger equation ( i.e. , of second order in @xmath2 , with @xmath3 the fermi velocity and @xmath4 the speed of light , or of second order in the fine structure constant @xmath5 ) . integrating out the fermions then leads to an effective theory for quantum magnets that generalizes and replaces the hertz - millis theory@xcite and its generalizations.@xcite more generally , the theory provides a derivation of spin - spin interactions in itinerant fermi systems in general , whether or not they are in a parameter regime where they develop long - range magnetic order . our results are therefore relevant , for instance , for fermionic atoms in optical traps or on optical lattices.@xcite this paper is organized as follows . in sec . [ sec : ii ] we consider , as a warm - up and to introduce various concepts , classical magnets , and show how the vector potential coupling to the magnetization gives rise to the dipole - dipole interaction . in sec . [ sec : iii ] we develop the technical machinery for dealing with quantum magnets and provide the derivations mentioned above . in sec . [ sec : iv ] we discuss our results and provide a summary and conclusion . some technical details are relegated to various appendices . we now proceed to derive an effective action for magnets that includes the effects of the fluctuating electromagnetic potential . we first consider the classical case as a warm - up ; we will generalize to the quantum case in sec.[sec : iii ] . consider a classical model for a ferromagnet with a three - component order parameter @xmath6 . in addition to the field @xmath7 we need to consider the electromagnetic vector potential @xmath8 , and the partition function @xmath9 is given by [ eqs:2.1 ] z & = & d[m , a ] e^s[m , a ] [ eq:2.1a ] + & & d[m ] e^-f[m]/t . [ eq:2.1b ] the model is defined by specifying the action @xmath10 , and in eq . ( [ eq:2.1b ] ) we have anticipated integrating out the vector potential to obtain an effective action @xmath11 in terms of the order parameter only . @xmath12 denotes the temperature , so @xmath11 is the free energy in mean - field approximation . throughout this paper we will use units such that boltzmann s constant and planck s constant are equal to unity , @xmath13 . for the order - parameter part of @xmath10 , we consider an @xmath14-symmetric @xmath15-theory , [ eqs:2.2 ] s_m = _ v dx . [ eq:2.2a ] @xmath16 represents a landau - ginzburg - wilson ( lgw ) theory of an isotropic ferromagnet with volume @xmath17 . the parameter @xmath18 contains the exchange interaction that leads to a magnetic ordering transition . in mean - field approximation this transition occurs at @xmath19 , with @xmath20 describing the paramagnetic phase , and @xmath21 the ferromagnetically ordered one . @xmath22 and @xmath23 are two additional model parameters , and @xmath24 . here , and throughout the paper , summation over repeated vector , tensor , and spinor indices is implied unless otherwise noted . note that @xmath16 is separately invariant under rotations in @xmath6 ( spin ) space and real space , respectively . the magnetization@xcite @xmath6 couples linearly to the curl of the magnetic vector potential @xmath25 : s _ = _ v dx m(x)((x ) ) , [ eq:2.2b ] with @xmath26 the bohr magneton in terms of the electron charge @xmath27 , the electron mass @xmath28 , and the speed of light @xmath4 . @xmath25 and @xmath29 transform as vectors in real space , and therefore @xmath30 is invariant only under co - rotations of spin space and real space . it is this coupling of the magnetization to the fluctuating vector potential that allows one to consider the magnetization as having a particular direction in real space . the vector potential is governed by s_a = _ v dx , [ eq:2.2c ] with @xmath31 any real number . the first term in eq . ( [ eq:2.2c ] ) is the magnetic energy , and the second term with coupling constant @xmath32 is a gauge fixing term . one popular choice is @xmath33 , which enforces a coulomb gauge , @xmath34 ; another one is the feynman gauge , @xmath35.@xcite either choice ensures a finite @xmath25-propagator . in coulomb gauge , it is a_i(k ) a_j(-k)= 4 t . [ eq:2.3 ] the vector potential can now be integrated out exactly , which leads to an effective action in terms of @xmath6 only . alternatively , we can consider the magnetic induction @xmath36 the fundamental field to be integrated out . in that case , the gauge fixing condition needs to be replaced by a constraint that enforces the maxwell equation @xmath37 . that is , the eqs . ( [ eq:2.2b ] ) , ( [ eq:2.2c ] ) , and ( [ eq:2.3 ] ) are replaced by@xcite s _ = _ v dx m(x)(x ) , [ eq:2.2b ] s_a = _ v dx _ 0 , [ eq:2.2c ] b_i(k ) b_j(-k)= 4t(_ij - k_i k_j ) . [ eq:2.3 ] either way we find [ eqs:2.4 ] & = & _ v dx + 2 ^ 2 _ k d_ij(k)m_i(k)m_j(-k ) , [ eq:2.4a ] with d_ij(k ) = k_i k_j . [ eq:2.4b ] the terms generated by integrating out the vector potential we recognize as the leading contribution to the dipole - dipole interaction@xcite plus a shift of the landau parameter @xmath18 by @xmath38 . the scalar potential @xmath39 , whose gradient is the electric field , does not lead to any magnetic interactions in a classical theory . this changes once the system is treated quantum mechanically , see sec . [ sec : iii ] below . the dipole - dipole operator @xmath40 , eq . ( [ eq:2.4b ] ) , transforms as a rank - two tensor in momentum ( or real ) space , and @xmath6 transforms as a vector in spin space . consequently , the dipole - dipole interaction is invariant under co - rotations in real space and spin space . this raises the question of other terms in the action that have the same symmetry properties . for instance , @xmath41 is allowed by symmetry . this term , and terms of higher order in the gradient , are generated by a renormalization of the action @xmath11 , as we now proceed to show . a renormalization of the action @xmath11 generates additional terms by virtue of the anisotropic @xmath42-propagator , which now reads m_i(k)m_j(-k)= + . [ eq:2.5 ] for instance , two - loop diagrams of the structure shown in fig . [ fig:2.1 ] -0 mm term.,title="fig:",width=113 ] both renormalize @xmath43 and lead to a new vertex _ k k_ik_jm_i(k)m_j(-k ) , [ eq:2.6 ] as well as to higher order anisotropic gradient terms . equation ( [ eq:2.6 ] ) represents the @xmath41 term that was mentioned above . conversely , if one starts with a theory that contains a @xmath44 term , which is allowed by symmetry and hence should be included in any landau theory , then the leading dipole - dipole term will be generated in perturbation theory even if it was not included in the bare action . the complete lgw action for a classical , isotropic heisenberg ferromagnet , up to terms quadratic in gradients and quartic in the order parameter , thus reads [ eqs:2.7 ] & = & _ v dx + & & - _ v dxdy m_i(x ) d_ij(x - y ) m_j(y ) + & & + _ v dx ( ( x))^2 . [ eq:2.7a ] here @xmath45 is the fourier transform of @xmath43 in eq . ( [ eq:2.4b ] ) , namely , d_ij(x - y ) = . [ eq:2.7b ] @xmath46 is the bare distance from the critical point in the effective lgw theory that takes into account the effect of the vector potential , and @xmath47 , @xmath48 , @xmath49 , and @xmath50 are the remaining landau parameters . as is clear from the above discussion , one expects the bare values of @xmath49 and @xmath50 to be small of order @xmath51 compared to the other parameters in natural units . these two parameters are usually set equal to zero in elementary treatments of classical heisenberg ferromagnets . the terms in the action so far are all even in the gradient operator , and hence invariant under spatial inversion . the spin - orbit interaction can eliminate this requirement by coupling the electron spins to the underlying lattice , provided the crystal structure is not inversion invariant . dzyaloshinsky@xcite and moriya@xcite ( dm ) showed that to linear order in the spin - orbit interaction @xmath52 the relevant term is _ = dx m(x ) ( ( x ) ) , [ eq:2.8 ] with @xmath53 . at a classical level , this term is purely phenomenological . dm showed how to derive it in the context of quantum mechanics , and in sec . [ sec : iii ] we will see how it arises in the context of field theory . we now turn to a quantum mechanical description of itinerant fermion systems in general , and certain types of magnetism in particular . we will show how the exchange interaction , the dipole - dipole interaction , and the dm interaction naturally arise in the context of a field - theoretic description of itinerant electrons . the former two lead to ferromagnetism , and the latter , if it is present , to helimagnetism . in appendix [ app : a ] we list the complete action for free quasi - relativistic electrons , to order @xmath51 , coupled to the electromagnetic field . several terms in the complete action are not relevant for our present purposes . the only effect of the darwin term , eq . ( [ eq : a.4 g ] ) , is to modify the coulomb interaction on length scales given by the electronic compton wave length @xmath54 . the relativistic mass enhancement , the last term in eq.([eq : a.4b ] ) , is a higher - order gradient term that is small compared to terms of the same form that are generated by renormalizing the final effective action . finally , the landau diamagnetic terms , eq . ( [ eq : a.4f ] ) , give rise to diamagnetism and , in the presence of an external magnetic field , landau levels . these effects are physically very different from ferromagnetism or helimagnetism , and we do not consider them here . finally , in quantum electrodynamics the fadeev - popov ghost field does not couple to any other fields . its only effect is to subtract the contribution of the unphysical longitudinal photon polarization to the free energy , and one has to keep it only if one is interested in the absolute value of the latter . neglecting all of these terms , we thus consider the following action : s[\|,;a _ ] & = & dx \|_(x ) _ ( x ) + & & -60pt + dxa_(x)a_(x ) + & & -60pt - ie dx ( x)n(x ) + dx b(x)_(x ) + & & -60pt + dx__1(x)__1 _ 2((x ) ) _ _ 2(x ) . + [ eq:3.1 ] here @xmath55 comprises real space position @xmath56 and imaginary time @xmath57 , and @xmath58 . @xmath59 and @xmath60 are grassmann - valued fields for electrons with spin projection @xmath61 , and the first term in eq.([eq:3.1 ] ) describes free electrons with chemical potential @xmath62.@xcite @xmath63 ( @xmath64 ; @xmath65 ) denotes the electromagnetic potential , with @xmath66 the scalar potential and @xmath67 the vector potential , and the second term in in eq . ( [ eq:3.1 ] ) describes the free electromagnetic field in feynman gauge ( i.e. , @xmath35 in eq . ( [ eq : a.4c])).@xcite note that both 4-vector potential fields in the second term carry covariant indices ; that is , the @xmath68-action is euclidian.@xcite @xmath69 and @xmath70 are the electronic number and spin density,@xcite respectively , with @xmath71 the pauli matrices , and the third term in eq . ( [ eq:3.1 ] ) describes the coupling of the electrons to the electromagnetic field , with @xmath72 the magnetic induction.@xcite finally , the last term in eq . ( [ eq:3.1 ] ) describes the spin - orbit interaction . note that both terms coupling the scalar potential @xmath73 to the fermions carry an extra factor of @xmath74 compared to what one might expect from the first quantized hamiltonian . this has the same origin as the euclidian metric mentioned above.@xcite equation ( [ eq:3.1 ] ) describes a continuum model . some of the effects we are interested in are present only in the presence of certain types of lattices , and we will comment later on the modifications that occur if the electrons are put on a lattice . the action , eq . ( [ eq:3.1 ] ) , depends only bilinearly on the electromagnetic potential . the latter can therefore be integrated out exactly , albeit at the expense of creating four - fermion terms . the latter represent electron - electron interactions that are mediated by the exchange of virtual photons . technically , we need the photon propagator , which we can read off the second term in eq.([eq:3.1 ] ) : [ eqs:3.2 ] a_(x)a_(y)= _ d(x - y ) , [ eq:3.2a ] with ( x - y ) = -4(_^2/c^2 + ^2)^-1(x - y ) , [ eq:3.2b ] or , in fourier space , ( k ) = 4/(_n^2/c^2 + k^2 ) . [ eq:3.2c ] here @xmath75 comprises a bosonic matsubara frequency @xmath76 and a wave vector @xmath77 . the result of integrating out the photons exactly is very complicated and involves interacting electronic modes in both the spin - singlet and spin - triplet channels , the particle - particle and particle - hole channels , and all angular - momentum channels . we will restrict ourselves to those terms that are most relevant for magnetism , i.e. interactions between spin - density fluctuations , or modes in the @xmath78-wave particle - hole spin - triplet channel . we organize the various contributions to the effective electron - electron interaction in powers of @xmath80 . to zeroth order only the term coupling @xmath73 to the number density @xmath81 in eq . ( [ eq:3.1 ] ) contributes . integrating out @xmath73 leads to a coulomb interaction [ eqs:3.3 ] s _ = - dxdy n(x)v_(x - y ) ( _ x - _ y ) n(y ) , [ eq:3.3a ] with v_(x ) = e^2/. [ eq:3.3b ] here we have neglected the dynamical nature of the photon propagator @xmath82 and have replaced it by its value at @xmath83 . the reason for this approximation is that fermi - liquid effects lead to a dynamical screening of the coulomb interaction that is a much larger effect than the relativistic dynamics inherent in eqs . ( [ eqs:3.2 ] ) . equation ( [ eq:3.3a ] ) contains number density fluctuations at all wavelengths . if one restricts the theory to interactions between long - wavelength fluctuations , then this interaction can be rewritten as a sum of parts that includes an interaction between spin - density fluctuations , see ref . and appendix [ app : b ] . the basic point is that an interaction between number density fluctuations at large wave numbers can be written as one between spin density fluctuations at small wave numbers . in an effective low - energy theory that contains only fluctuations at wave numbers smaller than some cutoff @xmath84 , eq . ( [ eq:3.3a ] ) therefore contains a contribution s _ = dx n_(x ) _ ( x ) , [ eq:3.4 ] where the prime on the integral indicates that only the small - wave - number contributions ( smaller than @xmath84 ) to the spin density @xmath85 are to be considered in order to avoid overcounting . as has been explained in ref . , it is convenient to choose the cutoff @xmath84 as a fixed fraction of the thomas - fermi screening wave number , and the spin - triplet interaction amplitude @xmath86 is a fermi - surface average over @xmath87 , with @xmath77 and @xmath88 pinned to the fermi surface . the restriction to small wave numbers will be understood from now on , and we will drop the prime on integrals . @xmath89 is the exchange interaction between electronic spin - density fluctuations that leads ferromagnetism . for later reference we note that @xmath86 is dimensionally an energy times a volume which , in this unrenormalized theory , is on the order of a rydberg times a fermi volume , or @xmath90 . we now turn to terms of @xmath91 . we first consider the vector potential @xmath25 , which couples to the spin density via the @xmath92 term in eq . ( [ eq:3.1 ] ) . since the coupling is directly to the spin channel , no phase space decomposition is necessary and integrating out @xmath25 proceeds as in the classical case , except that the @xmath93-propagator now is frequency dependent , see eqs . ( [ eqs:3.2 ] ) . we will comment on the consequences of this frequency dependence in sec . [ subsubsec : iii.c.2 ] below . neglecting the dynamical aspects of the dipole - dipole interaction for now , we obtain a contribution to the effective action s _ & = & 2 ^ 2 dx n_(x)_(x ) + & & 0 pt + dxdy ( _ x - _ y ) n_^i(x)d_ij(x - y ) n_^j(y ) , + [ eq:3.5 ] with @xmath40 from eq . ( [ eq:2.7b ] ) . the first term has the same form as the exchange interaction , eq . ( [ eq:3.4 ] ) , but is much smaller , as @xmath94 . the second one is the dipole - dipole interaction between the electron spins ; if one replaces the electronic spin density by its quantum mechanical and thermal average one recovers the classical dipole - dipole term in eq . ( [ eq:2.4a ] ) or ( [ eq:2.7a ] ) . we now return to the effects of integrating out the scalar potential @xmath73 . to @xmath91 the relevant contribution comes from the cross - term that multiplies the coulomb ( @xmath95 ) term and the spin - orbit ( last ) term in eq . ( [ eq:3.1 ] ) . contracting the two scalar potentials , integrating by parts , using eq . ( [ eq : c.2b ] ) , and keeping only terms that are bilinear in phase - space spin - density fluctuations , we obtain a contribution to the effective action s _ & = & _ ilm_ijk dxdy d(x - y ) \|__1(x)^l__1_4 + & & 0 pt ( _ _ 4(y ) ) ( \|__3(y ) ) ^m__3 _ 2 _ _ 2(x ) . + [ eq:3.6 ] after a fourier transform , eq . ( [ eq:3.6 ] ) can be written s _ & = & ( ) ^2 _ q , k , p d(q ) ( _ _ 1_4__3_2 ) . [ eq:3.7 ] here @xmath96 comprises a wave vector @xmath77 and a fermionic matsubara frequency @xmath97 , and @xmath98 and @xmath99 are used analogously . at this point we generalize to an effective interaction amplitude @xmath100 that depends on @xmath77 and @xmath88 in addition to @xmath101 . such a structure is generated in perturbation theory from the bare theory , where the interaction amplitude is simply given by the gauge field propagator , as we demonstrate in appendix [ app : d ] . we then have s _ & = & ( ) ^2 _ q , k , p ( _ _ 1 _ 4__3 _ 2 ) + & & _ _ 1(k - q/2 ) \|__3(p+q/2 ) _ _ 4(p - q/2 ) _ _ 2(k+q/2 ) . [ eq:3.8 ] hermiticity requires @xmath102 ( see also eqs . ( [ eqs : b.3 ] ) ) . we now employ the phase space decomposition explained in appendix [ app : b ] and focus on the large - angle scattering term , eq . ( [ eq : b.4b ] ) . projecting again on the spin density we obtain [ eqs:3.9 ] s _ = _ q d(q ) ( n_(q)_(-q ) ) , [ eq:3.9a ] where ( q ) & = & _ k , p g_kg_p ( k ) + & & + _ k , p g_kg_p ( k - p ) + & & d^(1)(q ) + d^(2)(q ) . [ eq:3.9b ] here g_k = ( _ k - ) [ eq:3.10 ] is a function that results from the projection onto the spin density and pins @xmath77 and @xmath88 to the fermi surface.@xcite equation ( [ eq:3.9a ] ) has the form of the result obtained by moriya.@xcite in what follows , we discuss the nature of the vector @xmath103 in some more detail . @xmath104 has the form [ eqs:3.11 ] ^(2)(q ) = i q^(2 ) + o(q^3 ) , [ eq:3.11a ] with @xmath105 a real vector given by ^(2 ) & = & _ k , p g_kg_p ( p - k ) . [ eq:3.11b ] @xmath106 can be written [ eqs:3.12 ] d^(1)_i(q ) = i d_ijq_j + o(q^3 ) , [ eq:3.12a ] with @xmath107 a real rank-2 tensor given by d_ij = _ k , p g_kg_p_ilmk_l p_me_j(k , p ) , [ eq:3.12b ] where _ i(k , p ) & = & dxdy(x_i - y_i ) e^-i(k+p)(x+y)/2 + & & -20pt dz ( ( k - p))d(x , y;z ) [ eq:3.12c ] with @xmath108 the fourier transform of @xmath109 in analogy to eq . ( [ eq : b.2b ] ) . @xmath107 has a symmetric part and an antisymmetric part . the latter can be combined with @xmath104 above to form a contribution to @xmath110 that we denote by [ eqs:3.13 ] ^(-)(q ) = iq + o(q^3 ) , [ eq:3.13a ] where d_i = d_i^(2 ) + _ k , p g_kg_p ( p_i k_j - k_i p_j)e_j(k , p ) . [ eq:3.13b ] the symmetric part can be written as a diagonal tensor plus a traceless rank-2 tensor , and the latter can be diagonalized by means of an orthogonal transformation that amounts to a spatial rotation . the symmetric part of @xmath107 we thus can write [ eqs:3.14 ] d^(+)_ij = d _ ij + a_i_ij [ eq:3.14a ] ( _ no _ summation convention ) where the @xmath111 obey _ i a_i = 0 . [ eq:3.14b ] if desirable , the @xmath111 can be explicitly constructed from eq.([eq:3.12b ] ) . we thus have a second contribution to @xmath110 that we denote by ^(+)(q ) = dq + i ( c a_x q_x + a_y q_y + a_z q_z ) + o(q^3 ) , [ eq:3.15 ] and ( q ) = d^(+)(q ) + d^(-)(q ) . [ eq:3.16 ] combining our results , and transforming back to real space , we now have [ eqs:3.17 ] s _ + s_ + s _ , [ eq:3.17a ] where s _ & = & d dx n_(x)(_(x ) ) , + [ eq:3.17b ] + s_ & = & dx n_(x ) , + [ eq:3.17c ] + s _ & = & dx ( _ ( x))(d_(x ) ) . [ eq:3.17d ] @xmath112 is the dzyaloshinsky - moriya interaction that is believed to be responsible for the helimagnetism observed in mnsi and fege.@xcite @xmath113 is a closely related term but absent in systems with cubic lattices ( e.g. , mnsi or fege ) due to the constraint , eq . ( [ eq:3.14b ] ) . finally , @xmath114 is another term that is allowed by symmetry , and is generated by the above derivation . all three of these terms are contained in moriya s general result,@xcite which takes the form of our eq . ( [ eq:3.9a ] ) above , but the effects of @xmath113 and @xmath114 have , to our knowledge , not been discussed explicitly . note that a necessary condition for any of these interactions to be nonzero is that the system is not invariant under parity : both @xmath106 and @xmath115 can be nonzero only if @xmath100 is odd under @xmath116 , or , equivalently , if @xmath117 is odd under @xmath118 . this implies that the dm interaction requires a lattice that is not invariant under spatial inversion ; in any continuum model , where the electron - electron interaction is necessarily even under parity , it vanishes . see appendix [ app : d ] and sec . [ sec : iv ] for further discussions of this point . we now have the following result . after integrating out the photons , the effective fermionic action reads s_[\| , ] & = & s_0 + s_ + s _ + s _ + s _ + s_ + s _ + & & s_0 + s_^. [ eq:3.18 ] here @xmath119 describes non - interacting electrons ( either free electrons or band electrons , depending on the model considered ) , and @xmath120 contains all interactions between modes other than the spin density , which we have not explicitly considered with the exception of the coulomb interaction , eqs.([eqs:3.3 ] ) . collectively we denote these two terms by @xmath121 . @xmath89 , @xmath122 , @xmath112 , @xmath112 , and @xmath114 are the exchange , dipole - dipole , and dzyaloshinsky - moriya interactions given by eqs . ( [ eq:3.4 ] ) , ( [ eq:3.5 ] ) , and ( [ eqs:3.17 ] ) , respectively . @xmath89 and @xmath122 are always present ; which , if any , of the terms in @xmath112 are nonzero depends on the details of the lattice structure and absence of spatial inversion symmetry is a prerequisite for any of them to be nonzero . collective , we denote the sum of these interactions in the spin - density or triplet channel by @xmath123 . for applications such as fermionic cold gases one will want to work directly with the fermionic action . for applications to magnets it is convenient to introduce a composite field @xmath124 whose expectation value is proportional to the magnetization . to this end we write [ eqs:3.19 ] s_^ = dxdyn_^i(x)_ij(x - y ) n_^j(y ) , [ eq:3.19a ] with _ ij(x - y ) & = & ( x - y ) _ + ^2(_x - _ y)d_ij(x - y ) + & & -60pt + ( x - y ) . + [ eq:3.19b ] here @xmath125 ( no summation convention ) . we now decouple eq . ( [ eq:3.19a ] ) by means of a hubbard - stratonovich transformation with a bosonic field @xmath124 . neglecting a constant contribution to the action , this allows us to write s_[\|,,m ] & = & s_0[\| , ] + & & - dxdy m_i(x)_ij(x - y)m_j(y ) + & & + dxdy . + [ eq:3.20 ] if one neglects the interacting part of @xmath121 this action depends only bilinearly on the fermion fields , and one can formally integrate out the fermions in order to obtain a theory entire in terms of the order - parameter field @xmath6 . this is a generalization of , and replaces , the hertz - millis theory.@xcite however , in general this is not a good strategy since it amounts to integrating out soft excitations , which means that any order parameter theory will in general not be well behaved . physically , these soft quasi - particle excitations can change the nature of the phase transition,@xcite or they themselves can become critical.@xcite in either case they must be treated on equal footing with the order parameter fluctuations . it therefore is technically advantageous , and physically more transparent , to work with the coupled field theory represented by eq . ( [ eq:3.20 ] ) . we conclude this section with a brief discussion of the dynamics of the dipole - dipole interaction , which we neglected in sec . [ subsubsec : iii.b.2 ] . if we restore the frequency dependence of the photon propagator and expand in powers of the frequency , then the leading dynamical contribution of the dipole - dipole interaction to eq . ( [ eq:3.20 ] ) takes the form s_^ = ^2 _ k ( _ ij - k_ik_j ) m_i(k ) m_j(-k ) . [ eq:3.21 ] as long as @xmath126 scales as @xmath127 , this scales the same as the @xmath128 term in the order - parameter theory that is induced by fermi - liquid effects,@xcite but has a prefactor that is smaller by a factor of @xmath129 . however , in classical dipolar magnets the order parameter is known to no longer be conserved.@xcite that is , @xmath130 scales as a constant , and this should be reflected in the quantum theory as well , although it is currently not known how this is realized . this suggests that the contribution shown in eq . ( [ eq:3.21 ] ) dominates the fermi - liquid - induced dynamics in a scaling sense , although it has a small prefactor , and will become important at sufficiently long time scales . we now discuss the significance of various interactions for a number of problems . as we mentioned in sec . [ subsec : iii.a ] , the energy scale for the exchange interaction in the bare theory is the atomic scale , or roughly @xmath131 . the corresponding length scale is on the order of @xmath132 . this is not consistent with the experimental fact that magnetic ordering is observed only at much lower temperatures ; e.g. , on the order of @xmath1 in fe and ni , on the order of room temperature in fege , and on the order of @xmath133 in mnsi . the reason for this discrepancy lies in the fact that the bare theory is renormalized in quantitatively substantial ways , and the corresponding energy and length scales in the properly renormalized theory are consistent with experimental observations.@xcite equations ( [ eq:3.5 ] ) and ( [ eqs:3.17 ] ) show that the dipole - dipole and dm interactions , within the framework of the bare theory , are weaker than the exchange interaction by a factor of @xmath129 , or about @xmath134.@xcite relative to the bare exchange interaction , this implies an energy scale on the order of @xmath135 , which is comparable with the ordering temperature in mnsi . on the other hand , the length scale associated with the dm interaction ( i.e. , the pitch length of the spin helix@xcite ) is only about @xmath136 ( in mnsi)@xcite to @xmath137 ( in fege),@xcite or only a factor of @xmath138 to @xmath139 larger than the atomic length scale . these observations indicate that there are strong renormalizations , due to band - structure and many - body effects , of all terms in the bare action , and that different terms are renormalized in different ways in different materials . while this makes it hard to make general statements , the bare theory suggests that the dm interactions and the dipole - dipole interaction are generically comparable in strength , and in mnsi , for instance , both are expected to be a substantial fraction of the ( greatly reduced by renormalizations ) exchange interaction . we thus conclude that there is no a priori reason to neglect the dipole - dipole interaction in any system where the dm interaction is known to be important . this calls for a re - evaluation of a number of interesting problems , some of which we list in the following subsection . we conclude by discussing a number of problems where the dipole - dipole interaction is either known to be important , or might be important , with an emphasis on low - temperature magnets and other fermion systems . \(1 ) classical heisenberg ferromagnets . this problem was worked on in great detail by aharony and fisher@xcite for the static critical behavior and by frey and schwabl@xcite for the dynamical critical behavior . the renormalization group done by aharony and fisher started from a nonlocal order parameter theory , eqs . ( [ eqs:2.7 ] ) , which leads to a somewhat nonconventional renormalization procedure . the nonlocality in eqs.([eqs:2.7 ] ) is due to integrating out the gauge field fluctuations or photons . it would be interesting to repeat this calculation starting from the coupled local field theory given by our eqs . ( [ eqs:2.2 ] ) before these fluctuations are integrated out . \(2 ) classical helimagnets . the standard phase transition treatment for helimagnetism due to the dm interaction is due to bak and jensen.@xcite neglecting the dipole - dipole interaction term they conclude that there is a fluctuation - induced first order phase transition from paramagnetism to helimagnetism . an interesting question is whether or not the dipole - dipole interaction modifies this conclusion . this seems especially relevant for mnsi where the phase transition is at low temperatures , i.e. , it is a weak helimagnet . \(3 ) it has been shown that in clean itinerant ferromagnets , the ferromagnetic transition is generically of first order at zero temperature.@xcite this conclusion ignores the effects of the dipole - dipole interaction terms . it would be very interesting to investigate if dipolar interactions can modify this generic conclusion . \(4 ) phase ordering is an important problem in ferromagnets.@xcite the dipole - dipole interaction terms has not been included in either the classical or quantum ( zero temperature ) ferromagnetic phase ordering problems . simple arguments indicate it will be important . \(5 ) fermionic cold atom systems . recently there has been a considerable amount of work on gases of fermions with dipolar interactions , see ref . and references therein . these systems are important for fermions in optical lattices . the dipolar interactions also serve as a mechanism for liquid crystal like phase formation in fermion systems . \(6 ) the dipole - dipole interaction term is important in the dynamics of classical antiferromagnets , both in the ordered phase , and near or at the phase transition if the systems is below its upper critical dimension.@xcite simple considerations suggest that they will also be important in low - dimensional ( 1 + 1 or 2 + 1 ) itinerant quantum antiferromagnets ; see sec . [ subsubsec : iii.c.2 ] above for one aspect of this problem . in this appendix we give the complete action for electrons interacting with electromagnetic fields in the weakly relativistic limit , up to and including terms of @xmath91 . let @xmath140 be the 4-vector potential , @xmath141 the pauli matrices , and @xmath26 the bohr magneton . we use standard relativistic notation , with covariant and contravariant indices related by a minkowski metric @xmath142 . expanding the dirac equation in powers of @xmath80 , one obtains , to order @xmath51 , the following hamiltonian in first quantization,@xcite [ eqs : a.1 ] = h _ + h [ eq : a.1a ] here @xmath143 is the pauli hamiltonian , _ & = & ( -i - a(x , t))^2 + e(x , t ) + & & - ( ( x , t ) ) , [ eq : a.1b ] and _ & = & ( ( x , t ) ) , [ eq : a.1c ] + h _ & = & ^2 ( x , t ) , [ eq : a.1d ] + h_m & = & ^4 . [ eq : a.1e ] describe the spin - orbit interaction , the darwin term , and the relativistic mass correction , respectively . via standard techniques,@xcite this theory can be reformulated in terms of an action that depends on fermionic ( i.e. , grassmann - valued ) field @xmath144 and its adjoint @xmath145 as well as the @xmath146-vector - potential field @xmath147 . for the partition function one obtains z = z_az _ , [ eq : a.2 ] where z_a = d[\|,]d[a _ ] e^s_a[\|,;a _ ] , [ eq : a.3 ] with an action [ eqs : a.4 ] s_a[\|,;a _ ] = s_[\| , ] + s_a[a _ ] + s_[\|,;a _ ] . [ eq : a.4a ] here s_[\| , ] & = & dx \|_(x ) _ ( x ) [ eq : a.4b ] describes the electrons with chemical potential @xmath62 , and s_a[a _ ] & = & dxa_(x)a_(x ) + & & -20pt + dx^2 [ eq : a.4c ] describes the electromagnetic fields , with @xmath148 a gauge fixing parameter . we use a 4-vector notation @xmath149 , @xmath150 for space and imaginary time . @xmath30 describes the coupling between the fermions and the electromagnetic field ; it contains four separate contributions : s_[\|,;a _ ] = s _ + s _ + s _ + s_. [ eq : a.4d ] here s_[\|,;a _ ] & = & -ie dx ( x)n(x ) + & & + dx b(x)_(x ) [ eq : a.4e ] is the coulomb and zeeman paramagnetic coupling that is included in the pauli equation , with @xmath151 and @xmath152 . s_[\|,;a _ ] & = & -i2dx_(x)a(x)_(x ) + & & + idx ( ( x))n(x ) + & & - dx a^2(x)n(x ) [ eq : a.4f ] is the landau diamagnetic coupling ; s_[\|,;a _ ] = dx(^2(x))n(x ) [ eq : a.4 g ] is the darwin term that , in the relativistic hydrogen atom , leads to the so - called _ zitterbewegung _ , and s_[\|,;a _ ] & = & dx__1(x)__1 _ 2 + & & ( ( x ) ) _ _ 2(x ) [ eq : a.4h ] is the spin - orbit coupling . the second factor in eq . ( [ eq : a.2 ] ) is [ eqs : a.5 ] z _ = d[\| , ] e^-s_[\| , ] [ eq : a.5a ] here @xmath153 is a one - component grassmannian field known as a fadeev - popov ghost field , with @xmath154 its adjoint , that is governed by an action s_[\| , ] = dx(x)_^(x ) . [ eq : a.5b ] for completeness , in this appendix we briefly recapitulate the arguments that lead to the generation of a spin - spin interaction , eq . ( [ eq:3.4 ] ) , from a density - density interaction , eqs . ( [ eqs:3.3 ] ) . for further discussion , see refs . , and , ; the latter also explain the relation to the work by shankar.@xcite consider an electron - electron interaction with an interaction amplitude @xmath155 . for simplicity we assume that the interaction is purely static , and translationally invariant , but otherwise general . the action has the form s & = & dx_1 dx_4 d w(x_1 -x_2 , x_3 - x_4 ; ( x_3 + x_4 - x_1 - x_2)/2 ) _ _ 1_2_3_4 \|__1(x_1,)\|__3(x_3 , ) + & & _ _ 4(x_4,)__2(x_2 , ) , [ eq : b.1 ] with @xmath57 a general rank-4 tensor . we define fourier transforms [ eqs : b.2 ] _ ( k ) = _ k e^-ikx w_k , p(q ) & = & dxdydz w(x , y;z ) , [ eq : b.2b ] with @xmath156 where @xmath77 is a wave vector and @xmath157 a fermionic matsubara frequency . hermiticity requires [ eqs : b.3 ] w^_k , p(q ) & = & w_k , p(-q ) , [ eq : b.3a ] + ^__1 _ 2 _ 3 _ 4 & = & _ _ 2 _ 1 _ 4 _ 3 . we then have [ eqs : b.4 ] s & = & ( ) ^2 _ k , p , q w_k , p(q ) _ _ 1_2_3_4__1(k - q/2)\|__3(p+q/2)__4(p - q/2 ) _ _ 2(k+q/2 ) [ eq : b.4a ] + & = & ( ) ^2 _ k , p , q w_(k+p - q)/2,(k+p+q)/2(p - k)__1_2_3_4__1(k - q/2)\|__3(p+q/2 ) _ _ 4(k+q/2 ) _ _ 2(p - q/2 ) + [ eq : b.4b ] + & = & ( ) ^2 _ k , p , q w_(p - k+q)/2,(k - p+q)/2(p+k)__1_2_3_4 \|__1(-k+q/2 ) & & _ _ 4(-p+q/2)__2(p+q/2 ) . [ eq : b.4c ] as long as all wave vectors are summed over , all three of these expressions are identical . if one restricts the summation in such a way that both @xmath158 and the modulus of the third argument of @xmath155 are smaller than a cutoff wave number @xmath84 , then we can represent the action as a sum of all three terms . they represent small - angle scattering , large - angle scattering , and @xmath159-scattering , respectively . alternatively , if one is interested in only one of these channels , one can pick the appropriate formulation of @xmath10 , restrict oneself to small wave numbers , and neglect the other channels . for our purposes , we are interested in the large - angle scattering channel , eq.([eq : b.4b ] ) . by choosing @xmath160 and making use of eq . ( [ eq : c.2a ] ) we obtain a term that has the structure of the exchange interaction , eq . ( [ eq:3.4 ] ) ( in addition to a contribution to the number - density interaction ) . the resulting spin - triplet mode is more complicated than a pure spin density , but it has an overlap with the spin density and can be restricted to the latter by the projection technique explained in ref . . note that a repulsive coulomb interaction results in an attractive exchange interaction due to a commutation of fermion fields that is necessary to write the result in the form of eq . ( [ eq:3.4 ] ) . similarly , by choosing @xmath161 and making use of eq . ( [ eq : c.2b ] ) we obtain the structure found in sec.[subsubsec : iii.b.3 ] ( in addition to terms that couple the number density and the spin density ) . here we give some properties of the pauli matrices that were used in sec.[sec : iii ] . let @xmath162 be the pauli matrices with the commutator property _ ijk^i^j = i^k ( i , j , k = 1,2,3 ) , [ eq : c.1 ] and @xmath163 the @xmath164 unit matrix . then the following identities hold : [ eqs : c.2 ] ^0__1 _ 2 ^0__3 _ 4 & = & ^0__1 _ 4 ^0__3 _ 2 + _ _ 1 _ 4__3 _ 2,[eq : c.2a ] + ^0__1 _ 2 _ _ 3 _ 4 & = & _ _ 1 _ 4 ^0__3 _ 2 + ^0__1 _ 4 _ _ 3 _ 2 + & & + _ _ 1 _ 4__3 _ 2 . [ eq : c.2b ] equation ( [ eq : c.2a ] ) is easily checked by a direct calculation , and eq.([eq : c.2b ] ) follows by multiplying eq . ( [ eq : c.2a ] ) by @xmath165 from the right and using eq . ( [ eq : c.1 ] ) . in this appendix we demonstrate that a general interaction amplitude of the type used in eq . ( [ eq : b.1 ] ) is generated in perturbation theory from an ordinary two - body interaction . for simplicity , we consider spinless fermions interacting via an interaction potential @xmath166 ; spin dependence or gradients , as in the last term of eq . ( [ eq:3.1 ] ) , are easily added . we also consider a translationally invariant model ; we will comment on this feature below . the bare electron - electron interaction is described by a term in the action s = dxdy v(x - y)\|(x)\|(y)(y)(x ) . [ eq : d.1 ] our goal is to construct an effective interaction of the form s _ & = & dx_1 dx_4 + & & w(x_1-x_2,x_3-x_4;(x_3+x_4-x_1-x_2)/2 ) + & & ( x_1 ) \|(x_3)(x_4 ) ( x_2 ) [ eq : d.2 ] that has a structure necessary for contributing to the tensor @xmath107 , eqs.([eq:3.12b ] , [ eq:3.12c ] ) , see fig . [ fig : d.1 ] . the bare interaction corresponds to w^(1)(x , y,;z ) = ( x)(y)v(z ) . [ eq : d.3 ] this does not contribute to @xmath107 since it enforces @xmath167 . more generally , contributions to @xmath155 with the property @xmath168 do not contribute to @xmath107 . at second order in @xmath166 , there are several diagrams , both tree - level and one - loop diagrams , that have this property . consider , however , the diagrams shown in fig . [ fig : d.2 ] . in order for @xmath155 to contribute to @xmath107 we must have @xmath170 , see eq . ( [ eq:3.12c ] ) . from eq . ( [ eq : d.4 ] ) we see that this is the case if and only if @xmath171 has an odd component @xmath172 . as far as the contribution to @xmath107 is concerned we can thus replace @xmath166 in eq . ( [ eq : d.4 ] ) by @xmath173 , and this automatically ensures @xmath174 . if @xmath166 is the propagator of a scalar field , such as the quantity @xmath82 in sec . [ sec : iii ] , then these symmetry properties can obviously not be realized . however , in a realistic solid - state model @xmath166 represents the screened coulomb interaction , @xmath175 , and the dielectric function @xmath176 has a contribution from the lattice in addition to an electronic contribution . the latter will only have the symmetry of the space group of the lattice , and on a lattice without inversion symmetry one will have @xmath177 . translational symmetry will also be broken , of course , but this is not necessary for making the dm interaction nonzero , as the above example shows . note that one can consider a coarse - grained continuum theory for the dm interaction , see eq . ( [ eq:2.8 ] ) , but the relevant landau coefficient _ must _ depend on the underlying lattice and vanish if the true continuum limit is taken . we would like to thank achim rosch and thomas vojta for discussions . this research was initiated at the aspen center for physics , and supported by the national science foundation under grant nos . dmr-09 - 29966 and dmr-09 - 01907 .	it is shown how the exchange interaction , the dipole - dipole interaction , and the dzyaloshinsky - moriya interaction between electronic spin - density fluctuations emerge naturally from a field - theoretic framework that couples electrons to the fluctuating electromagnetic potential . semi - quantitative estimates are given to determine when the dipole - dipole interaction , which is often neglected , needs to be considered , and various applications are discussed , with an emphasis on weak ferromagnets and on helimagnets .

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.